Working Paper Series

The evolution of gender and racial occupational segregation across formal and non-formal labour markets in Brazil – 1987 to 2006 Paola Salardi

ECINEQ WP 2012 – 243

ECINEQ 2012 – 243

January 2012

www.ecineq.org

The evolution of gender and racial occupational segregation

across formal and non-formal labour markets in Brazil – 1987 to 2006*

Paola Salardi†

University of Sussex

Abstract This paper provides a unique analysis of the evolution of gender and racial occupational segregation in Brazil from 1987-2006. Drawing on a novel dataset, constructed by harmonizing national household data over twenty years, the paper provides extensive new insights in the nature and evolution of occupational segregation over time, while also providing important new insights into the forces driving these changes. The results presented here expand upon existing research in the developing world in several directions. First, the new dataset constructed for this study allows the analysis to cover a longer time period than has previously been possible. Second, the analysis explores both gender and racial segregation side by side. Third, all of the analysis is conducted for the labour market as a whole, and disaggregated into the formal, informal and self-employed labour markets. Fourth, the paper decomposes the key driving forces that lie behind trends in occupational segregation. The paper presents three major findings: first, gender segregation is always considerably greater than racial occupational segregation, but racial segregation has been more persistent over time and has several features that make it comparatively worrisome; second, while occupational segregation is declining by both gender and race, the decline has been greater in the formal labour market. Third, the decomposition of segregation measures over time reveals that changes in the internal gender and racial composition of occupations have driven improvements over time. These important differences between formal and non-formal labour markets provide preliminary insights into the possible importance of formal labour market policies and institutions in shaping outcomes. Keywords: Brazil, Gender, Race, Occupational Segregation, Informality. JEL Classification: J15, J16, J71, O17, O54.

* This paper draws on two chapters of my PhD thesis. I thank Gustavo Bobonis, Jairo Castro, Richard Dickens, Patricia Justino, Julie Litchfield, Rafael Osorio, Wilson Prichard and Barry Reilly and participants at the 2011 AIEL conference in Milan and at 2011 PhD conference at the University of Sussex for useful comments and suggestions. All errors are my own. † Contact details: p.salardi@sussex.ac.uk.

2  

1. Introduction

Occupational segregation represents one of the core themes in the labour economics

literature (e.g., Anker 1997, Reardon and Firebaugh 2002, Fryer 2010, King 1992, Charles and

Grusky 1995 and Watts 1998), but despite the centrality of occupational segregation to any

understanding of labour market outcomes, studies of occupational segregation have been

surprisingly rare in developing countries. This paper seeks to address this gap in the literature by

providing the most detailed existing analysis of the evolution of gender and racial occupational

segregation in Brazil, covering the years from 1987-2006.

One reason for the absence of such studies in developing countries has been the lack of

sufficiently detailed and complete data. The first empirical contribution of this study

correspondingly lies in the introduction of a harmonized reclassification of occupations drawn

from the Brazilian national household surveys (the Pesquisa Nacional por Amostra do

Domicilios (PNAD)) from 1987 to 2006. This new data set addresses the absence of a

systematic and consistent classification of occupations over time in the PNAD, and makes it

possible to explore labour market trends over a longer period than has previously been possible,

while using categories that also facilitate international comparison.

Having constructed this new harmonized occupational classification, this paper

contributes to our understanding of occupational segregation in Brazil, as well to the broader

literature, in several ways. First, drawing on the new data, it assesses changes in Brazilian

occupational structure, and the magnitude of occupational segregation,1 over time, providing

more accurate and complete findings by employing several alternative measures of occupational

                                                        1 In this study we focus on horizontal (or nominal) occupational segregation, which captures disparities in occupational attachment. That is, it captures the extent to which different population groups are over represented in certain occupations, and relatively absent from others. This is distinct from vertical (or hierarchical) occupational segregation, which captures disparities in the positions (and thus the pay) received by different population groups within individual occupations (Semyonov and Jones 1998, Blackburn et al 2001, Blackburn and Jarman 2005). That is, vertical segregation captures the extent to which certain population groups may be overrepresented in more senior or high paying positions within individual occupations. It is important to be extremely clear about definitions, as these definitions of horizontal and vertical segregation sometimes differ within the economics literature. Finally, it is important to note the distinction made by Merkas and Anker (1997): if horizontal segregation refers to differences in employment across occupations, and vertical segregation refers to different positions within occupational groups, then when there is a sufficiently detailed number of occupations used in the analysis, these two concepts tend to become equivalent.

3  

segregation. Second, the analysis addresses both gender and racial2 segregation side by side, in

order to highlight commonalities and differences in both levels and trends over time. Third, all

of the analysis is conducted for the labour market as a whole, and disaggregated into the formal,

informal and self-employed labour markets, highlighting important differences and the potential

importance of formal labour market policies and institutions in shaping outcomes. Fourth, the

analysis is also conducted disaggregated by several key characteristics of the labour force, in

order to identify specific demographic, educational, sectoral and spatial patterns. Finally, the last

section employs a technique proposed by Deutsch et al (2009) in order to decompose the causes

of declining segregation into, respectively, changes in the gender or racial composition within

individual occupations, changes in the overall occupational structure and changes in the sub-

population shares of the entire workforce.

A central goal of the paper is to not only describe trends over time, though this is

valuable in its own right, but also to gain preliminary insights into the driving forces behind

these trends. This paper nonetheless makes two novel contributions in this direction. First, by

disaggregating the analysis into the formal and non-formal labour markets, we gain indirect

insight into the possible impact of anti-discrimination legislation (ADL) and other labour market

policies and institutions. The formal sector provides scope for regulated labour markets to

function, and as such different outcomes with respect to gender and racial occupational

segregation across the formal, informal and self-employed sectors are likely to reflect the impact

of this regulation. While the findings are only indicative, the evidence is consistent with the

view that labour market rules have contributed to lower levels of occupational segregation, as

segregation has decreased faster in formal labour markets. Second, the decomposition

methodology employed in the final section of the paper finds that it is changes in the internal

composition of occupations that have driven improvements over time among all sectors, rather

                                                        2 We employ the commonly recognized and understood term “racial” segregation to denote segregation based on differences in skin colour. However, while we employ this term for the sake of simplicity, the term “skin-colour based” segregation is arguably more accurate, as the Brazilian population is generally held not to be classifiable into ethnicities. Brazil is a multi-racial society, among which the dominant population groups have historically included brancos (whites), pretos (black people), amarelos (Asians), and pardos (brown people, including mulatos, caboclos, cafuzos, mamelucos and mestiços), while since 1992 indigenous people’s have been split from pardos and considered an independent category. Given this complexity, we adopt Telles’ approach and consider all non-white populations together, in part because the distinction between the black and brown categories is somewhat subjective, and frequently reflects the perceptions of white individuals (Telles 1992, Telles and Lim 1998). In addition, the white and non-white categorization suits our approach due to the need for a binary variable when employing standard methodologies for occupational segregation analysis.

4  

than declining segregation simply being driven by the entry of women and non-whites into the

labour force. Moreover, across both gender and racial segregation, changes in the occupational

structure have contributed to increasing levels of segregation, though this effect is concentrated

entirely in the non-formal labour markets.

The paper is structured as follow. Section 2 reviews existing studies that investigate

occupational segregation and informality, particularly in the Brazilian context. Section 3

describes the construction of the new data set, based on the harmonized classification of

occupational codes over twenty years. Section 4 provides a brief overview of key changes in the

Brazilian occupational structure over time. Section 5 presents the analysis of gender and racial

occupational segregation by applying several well-known segregation measures. Section 6

presents the decomposition of changes in both gender and racial occupational segregation over

time, while the final section concludes.

2. Literature Review

While occupational segregation is among the core topics of labour economics

surprisingly little systematic empirical research has been conducted on this topic in developing

countries, while few have sought to carefully compare gender and racially based segregation, and

none have drawn clear connections between informality and broader trends in occupational

segregation. A careful review of the literature reveals 25 studies that measure occupational

segregation in individual countries, of which the majority focus on gender occupational

segregation3 and only three look at developing countries.4 The only studies to look jointly at

gender and racial segregation focus on developed countries (Albelda 1986, King 1992 for U.S.

and Neuman 1994, 1998 for Israel). In addition to these individual country studies, a smaller

                                                        3 For U. S. see Albelda (1986), Blau and Hendricks (1979), Baunauch (2002), Cotter et al (2003), Hutchens (1991, 2004), King (1992) and Watts (1995); for United Kingdom see Hakim (1992, 1993) and Watts (1998); for Australia see Lewis (1982), Moir & Selby-Smith (1979) and Karmel & Maclachlan (1988); for Ireland see Reilly (1991); for Israel see Neuman (1994, 1998), for Switzerland see Deutsch et al (1994) and Fluckiger and Silber (1999), for Brazil see Oliveira (2001), for Spain see Mora & Ruiz-Castillo (2003), for Mexico see Calonico and Nopo (2007), for Colombia see Castro and Reilly (2011). Among previously cited works, Albelda (1986), King (1992) and Neuman (1994, 1998) also explore racial occupational segregation. In addition we find two studies only on racial segregation in the US, such as Boisso et al (1994) and Maume (1999). 4 Calonico and Nopo (2007) on Mexico, Castro and Reilly (2011) on Colombia, Oliviera (2001) on Brazil.

5  

number of studies adopt a cross-country perspective (Blackburn et al 1993, Charles and Grusky

1995, Melkas and Anker 1997, Semyonov and Jones 1999, Anker et al 2003, Deutsch et al 2005

and Deutsch and Silber 2005), of which only two consider experiences in developing countries5.

Focusing specifically on Brazil, theoretical and empirical research looking at wage

discrimination is extensive (see among others, Soares 2000, Arcand and Hombres 2004, Arias et

al 2004, Arabsheibani et al 2003, Carvalho et al 2006), but the only empirical study of

occupational segregation is, to the best of the author’s knowledge, Oliveira’s (2001) study

finding that gender occupational segregation, measured by the Duncan Index, declined by three

percentage points between 1987 and 1999 when measured at the 3-digit level of occupational

classification. There is a clear opportunity to update Oliveira’s work over a longer period of

time, and to move beyond it in looking at racial segregation, and in disaggregating the analysis

into the formal, informal and self-employed sectors.

The decision to disaggregate the analysis into the formal and non-formal sectors reflects

the greater risk exposure faced by those in the informal sector, and the potential to highlight the

factors contributing to different patterns of occupational segregation across sectors. Given this

focus on informality, it is useful to briefly review existing research looking at the evolution of

the Brazilian informal sector.

Arriving at a precise definition of informality is challenging both conceptually and

empirically. The first conceptualization of informality is attributable to Hart (1971, 1972) and it

mainly used to the concept of informality as small business mostly characterized by rudimentary

technologies. Conceptually, the informal sector can be seen in productive terms, as offering

employment to micro-entrepreneurs, families engaged in small businesses, precarious and

unskilled workers, or it can be viewed though a more legalistic lens, as a site for unregulated and

illegal activities that evade taxation (Gasparini and Tornarolli 2007). These two definitions

imply a need to distinguish between informality and the shadow economy, that is, between small

businesses and other illegal and tax-avoiding activities (Cacciamali 1982, 1983).

Alongside these conceptual issues is the question of how, methodologically, to identify

informal workers. Some early studies of Brazilian informality focused on wage workers without

labour contracts, self-employed individuals, employers earning up to a certain portion of the

                                                        5 Deutsch et al (2005) on Costa Rica, Ecuador and Uruguay, and Anker et al (2003), who analyze cross-country variation in occupational segregation, including both developed and developing countries.

6  

minimum wage, unpaid family workers, and domestic service workers (see Jatoba 1987 and

Gatica 1989 cited in Carneiro 1997), while other studies adopted a definition of informality

based on the payment of social security contributions (see Cacciamali 1988, Telles 1992). In a

more recent paper, Henley et al (2009) compare three different definitions of informality: i)

contract status, based on the possession of a signed labour card; ii) social security status, based

on contribution to a social security institution; and iii) formal sector activity, based on

employment within a firm with more than five employees. They find that only 40% of cases are

classified as informal across all three definitions of informality, highlighting the importance of

clear definitions.

Despite these definitional challenges, most empirical studies in Brazil have defined

informal workers as those workers without signed work cards, the carteira de trabalho (Carneiro

1997, Soares 2004, Ulyssea 2006). However, even this definition leaves important issues

unresolved. The first relates to the treatment of the self-employed, as different studies have

opted to exclude them, include them in the informal sector or treat them as a separate component

of non-formal labour markets, as is the case here (Maloney 2004, Almeinda and Carneiro 2007).

The second related to employers, for whom identifying formality or informality is very difficult,

as they do not possess the carteira de trabalho. The ILO considers employers with less than five

employees to be in the informal sector, and this definition is adopted by Bosh et al (2007) in their

study of the Brazilian informal sector. However, this threshold varies from country to country

(see the discussion in Bosh et al, 2007), information on the number of employees might be

missing and some small firm employers may be formal according to other metrics, such as the

payment of social contributions.

While there are both conceptual and methodological issues related to establishing a

universal definition of informality, there is consensus that the Brazilian informal market is large,

with estimates placing it at 50% of more of the total labour force (e.g. Urani 1996, Carneiro

1997, Soares 2004, Bosch et al 2007, Ramos and Ferreira 2005). Much of this literature further

argues that the informal sector has been rising over time, with Bosch et al (2007) in particular

arguing that it has increased by 10% during the 1990s. However, Ramos and Ferreira (2005)

argue that this worrisome increase in informality in Brazil is primarily confined to urban areas

and the manufacturing sector, while attention to the nation as a whole provides a more mixed

pattern. Equally important, the recent work of Bosch et al (2007) argue that the informal sector

7  

is closely intertwined with the formal sector labour market, though providing an attractive

alternative for more flexible and unregulated business outside of government regulations (see

also Carneiro 1997, Fiess et al 2008). Consistent with this view, they find that both formal and

informal labour markets are highly pro-cyclical and strictly interrelated, as most transitions from

the formal to the informal sector occur within particular industries, rather than resulting

primarily from structural changes in the importance of different economic sectors (see also

Maloney 1999, 2004).6

Along with capturing broad trends, several authors seek to explain the determinants of

these trends, though disentangling causality is inherently difficult. Bosch et al (2007) conclude

that the primary driver of the growing informal sector that been institutional reforms affecting

the labour market, including union power, firing costs and overtime rule, while trade

liberalization has a played only a very minor role. Several other authors broadly echo this

finding, arguing that stronger enforcement of labour protection in the formal sector may reduce

hiring and encourage informality (Carneiro 1997, Holk 2002, Ulyssea 2010). Goldberg and

Pavcnik (2003) similarly find no impact of trade variables on trends in the informal sector, while

Paes de Barros and Corseuil (2001) argue, contrary to Bosch et al (2007), that there is no

evidence that labour market regulations have driven changes in levels of informality. Ultimately,

these studies generally find that tighter regulation in the formal sector is an important reason for

informality. On the other hand, they also do not address other factors that may also shape trends

in the informal sector, such as reforms in the public health sector or the huge migration of

workers from rural areas to urban/metropolitan areas, primarily in the South-East of Brazil.

Finally, and of greatest interest to this paper, several studies have highlighted key features

of the informal sector that appear to reflect an important connection between informality and

occupational segregation. Telles (1992) and more recently Ulyssea (2010) note that the informal

sector continues to be dominated by women and non-whites, and particularly by non-white

women, though they both note the comparative lack of attention to issues of race in studies of the

Brazilian informal sector. These patterns are most clearly revealed in the experience of domestic

service workers, who are primarily women, disproportionately non-white and remain largely

informal (Abramo 2004). Finally, Telles (1992) notes that educational attainment plays a

                                                        6 The limited role of structural changes in justifying the explosion of the informal sector is also acknowledged in Ramos and Ferreira (2005).

8  

particularly important role in the ability of women to enter the formal labour market, with

uneducated women overwhelmingly employed in the informal market, while the same pattern

does not hold true for men. These very cursory findings from the literature provide an initial

inspiration for the decision here both to consider gender and racially based segregation side by

side and to consider the formal and non-formal labour markets separately.

3. The Construction of the new Classification of Occupational Codes

Most studies of occupational structure, occupational segregation and informality in Brazil

are based on the Pesquisa Nacional por Amostra do Domicilios (PNAD), but these studies are

plagued by the existence of a major break in the data on occupations, owing to a radical change

in the way that occupations have been classified especially since 2001. The result is that it has

been difficult or impossible to conduct studies of the evolution of occupational segregation in

Brazil over a protracted period of time and including recent developments. 7 In order to

overcome this problem, this study is based on a novel re-classification and harmonization of the

occupational codes from successive PNAD surveys, thus allowing for the analysis of trends in

occupational structure and segregation over a longer period, and in greater detail, than has

previously been possible. The most detailed and compatible existing re-classification of

occupational codes using the national household survey (PNAD) is attributable to Osorio (2008),

who constructed 46 occupational codes at the 2-digit level over the period 1986 to 2006.

Nonetheless, our effort is more accurate than previous re-classification efforts, including those

studies that have employed different surveys, for example, Barros et al (1997), Lovell (1994,

2000, 2006), Lago (2006). 8

                                                        7 While Oliviera (2001) considers occupation segregation over a relatively protracted period, from 1981 to 1999, this was only possible because the study focused on years prior to the radical change in the PNAD classification of occupational codes (Oliveira, 2001). The same is true for Machado et al (2003), who look at trends over the period 1981-2001 (again stopping before the break in the PNAD data in 2001) by aggregating over 300 occupational codes at the 3-digit level into 67 groups at 2-digit level. 8 The Brazilian national commission for classification, the Commisão Nacional de Classificação (CONCLA) has also prepared a re-classification that recoded the official national classification of occupations (CBO-94), and the official classification of occupations used by the Census, in order to make both compatible with the international classification standard ISCO-08 (also discussed in Muendler et al, 2004). However, the official national classification of occupations by CONCLA does not address the disruption of the time series by changes in the occupational codes for the PNADs starting in 2002.

9  

The methodology for this reclassification is described very briefly here, due to the lack of

space.9 The key challenge in this reclassification has been to ensure sufficient homogeneity

within each occupational grouping. To this end, key features of each occupational group have

been analysed and taken into account, and we have paid special attention to the mean of main job

earnings and both the means and modes of the maximum level of education attained. A key

element in understanding this process, which is stressed in Muendler et al (2004), is that in re-

classifying occupations, we have transformed the more profession-based Brazilian classification

system CBO-94 into the more skill-oriented international system ISCO-08. The resulting re-

classification of Brazilian occupational codes is consistent over time and compatible with

international standards, containing 80 occupational codes at 3-digit level, 26 occupational codes

at 2-digit level and 10 occupational codes at 1-digit level (see the complete re-classification in

Table A1 in the appendix). The major occupational groups at 1-digit level are, in order:

Legislators, senior officials, and managers; Professionals; Technicians and associate

professionals; Clerks; Service workers and shop and market sales workers; Skilled agricultural

and fishery workers; Craft and related trades workers; Plant and machine operators and

assemblers; Elementary occupations, and; Armed forces.

4. Background: The Evolution of Brazilian Occupational Structure

While our focus is on the evolution of occupational segregation over time, it is useful to

begin with an overview of broad changes in occupational structure, as well as how these new

estimates compare to earlier research. What follows focuses, in particular on highlighting

differences based on gender and race, as well as across the formal, informal and self-employed

sectors. Throughout the analysis, we consider the entire national labour market, i.e. all five

regions of Brazil and both urban and rural areas. As noted earlier, this is important, as conditions

are somewhat variable across the country and, as such, analysis of specific regions or

metropolitan areas risks capturing trends that do not reflect the situation of the entire nation. Of

course, this national focus risks obscuring important differences across sub-groups or regions,

and the analysis thus concludes with a discussion of these differences.

                                                        9 Additional information is, of course, available from the author.

10  

As was already noted, we divide the entire labour market into the formal, informal and

self-employed sectors. The formal sector comprises private sector employees with signed labour

cards, domestic workers with signed labour cards and civil servants. The informal sector

comprises private sector employees and domestic workers without a signed labour card. Given

that our sample covers both urban and rural areas, agricultural workers both with and without

signed labour cards are included in the formal and informal sectors respectively. We choose to

keep self-employed workers separate from informal workers due to differences in composition

and trends amongst these two non-formal sectors. We exclude military forces and employers.10

Finally, our sample excludes workers who are not remunerated or for whom the wages variable

is missing. As the exclusion of ‘zero wage’ observations is likely to result in an underestimate of

the non-formal sectors, we perform a robustness check to see if accounting for these zero wages

observations has a significant effect on our estimates of informality and segregation.

4.1 Distribution of Workers between the Formal and Non-Formal Sectors

Dividing the Brazilian employed labour force into the formal, informal and self-employed

sectors reveals two broad messages. First, the informal and self-employed sectors cover more

than half of the entire sample across all twenty years. Second, the distribution of workers across

these three sectors has remained nearly constant over time (Figure 1). The formal sector has

increased by only 1.23 percentage points during the last two decades, moving from 45.53% in

1987 to 46.76% in 2006. On the other hand, both the informal and self-employed sectors have

declined by 0.6 percentage points. The absence of an increase in informal sector activities at the

national level is in line with previous research by Ramos and Ferreira (2005). While several

other studies have reported rising informality, this rise, as generally acknowledged in several

empirical studies,11 is a more restricted phenomenon that refers mainly to private employees in

metropolitan areas, especially in the South-East region. Part of the explanation for the

differential trend when looking at the national level is the fact that our sample accounts for

                                                        10 While we could have followed Bosh et al (2007) in using the ILO threshold to distinguish formal and informal sector employers (with those with less than five employees considered informal), this method appears to be problematic in this case, as the data reveals that at least 50% of small firm employers pay social security contributions and, as a consequence, should not be considered informal workers. 11 See, for example, Carneiro (1997) and Bosh et al (2007). Carneiro (1997) report a rise of informality by looking at the metropolitan area of Sao Paulo. Bosh et al (2007) claim that the informal sector has increased by 10 percentage points between 1985 and 2002 by considering only private sector in six metropolitan areas.

11  

agricultural and domestic workers, both of whom have experienced an increase in the “degree of

formalization” of their occupations over time.12

[Figure 1 about here]

Turning to trends by gender and race, Figure 2 reveals that although male and white

workers have traditionally dominated the Brazilian labour market, the presence of women and

non-white workers has consistently increased over time, by 5.74 and 6.59 percentage points

respectively. Looking specifically at women, despite the huge increase of women in the entire

labour market, they are still generally underrepresented in Brazilian labour market and are more

present in the informal sector (Wajman and Rios Neto 2000, Soares and Izaki 2002, World Bank

2002a, 2002b). In the case of race, the increasing share of non-white workers in the labour force

has resulted in their exceeding the share of white workers starting from 2003. Despite this rapid

increase in the share of women and non-white individuals in the labour market, it is important to

note that both groups continue to be strongly overrepresented in the informal sector relative to

formal sector employment.

[Figure 2 about here]

While women have entered the labour force rapidly, the experiences of white and now-

white women have been very different (Figure 3). White women have overwhelmingly joined

the formal sector, with their participation in the informal sector actually declining, while non-

white women continue to be sharply overrepresented in the informal sector. Along the same

lines, we see an increasing representation of non-white workers in both non-formal sectors,

mainly driven by non-white women in the informal sector (roughly six percentage points) and

non-white men in the self-employed sector (roughly five percentage points).

[Figure 3 about here]

                                                        12 In our dataset, the share of formal agricultural workers increased from 12% to 18% between 1987 and 2006, while formal domestic workers increased from 19% in 1992 to 27% in 2006 (see Fonseca & Rayp (2011) on the formalization of agricultural workers and Berg (2010) and ILO (2010) on the formalization of domestic workers).

12  

4.2 Labour Market Trends, Disaggregated by Characteristics of the Labour Force

In addition to these broad trends, more nuanced messages emerge by considering

different populations groups disaggregated by key characteristics (age, education, economic

sector, region, and urban/rural).13 Due to the constraints of space we selectively highlight the

most important insights that emerge, rather than presenting each individual piece of the analysis.

Looking first at employed female labour force, we find growing participation among

young and adult women with respect to their male peer group, but also find that the share of

elderly women in the labour force has risen rapidly, as women now remain in the labour market

longer (Wajnman et al, 2006). With respect to years of education, women have always

represented more than half of the total educated labour force: well educated women are much

more likely to be engaged in the labour force. The presence of women is predominantly in the

tertiary sector and especially in the social services, trade and hospitality and financial services

sectors. Finally, female participation has increased primarily in the South and South-East

regions, and primarily in urban areas.

Non-white workers have increased across all age groups, and particularly among young

workers, with young non-white workers becoming more than half of the labour force after 2000.

Non-white workers represent the predominant share of the illiterate labour force, at an average of

70%, but their share of the more educated workforce is also, encouragingly, increasing.14 The

presence of non-white workers is predominantly in the primary sector, followed by the secondary

sector, while non-white participation is increasing homogeneously across all three sectors over

time.15 The North and North East regions record the largest share of non-white workers, while

non-white workers also represent the majority of the rural labour force over time. The fact that                                                         13 We define three main age groups: young (aged 15-29), adult (aged 30-49) and elderly (aged 50-65). For educational attainment the employed labour force is divided among illiterate workers, workers that achieved compulsory school level only and more educated workers with more than a compulsory school degree. We consider a standard three sector grouping of economic activities (primary, secondary and tertiary sectors) as well as a detailed breakdown of different economic activities: a) agricultural, forestry and fishing activities (hunting is included as well), b) mining, c) manufacturing, d) services related to electricity, gas and water provision, e) construction, f) trade activities and services related to hospitality and tourism, g) transport and storage activities, h) financial services (including insurance services) and real estate and, finally, i) social services (including health and education). Finally, with respect to geography we consider the five main regions of Brazil (North, North-East, South-East, South and Central-West) as well as the division between urban and rural areas. 14 Among those that have attained only compulsory school, the share of non-white workers has increased from 47% to 59%, while their share of the workforce with more than compulsory education has similarly increased, from 20% up to 34% (the rise of educational attainment among non-white Brazilian is also documented in Osorio, 2008). 15 At a more detailed level of disaggregation, non-white workers are most heavily represented in agricultural activities and the construction and mining sectors (mainly non-white men) followed by trade and hospitality and social services (mainly non-white women).

13  

the share of non-white workers in the North and North-East has remained relatively constant,

despite the aggregate rise of non-white participation in the labour market, may be evidence of

migration across regions during this period.16

4.3 Trends in Occupational Distribution

If we turn to looking at the occupational distribution – that is, investigation of the

professions in which members of different population groups are primarily employed – a number

of further insights emerge. First, although there has been an increase in female and non-white

participation in the labour market, the occupations in which these groups are primarily employed

have, somewhat surprisingly, remained relatively stable over time (Figure 4). That is, a large

part of the increase in female labour force participation has been directed at the same economic

sectors in which they have historically been employed, primarily in the tertiary sector. For non-

white workers we see somewhat greater change in the occupational distribution, with significant

movement into the secondary and, particularly, tertiary sectors, but here too this movement has

been largely into areas in which that have traditionally been employed. The result is that new

female and non-white workers have largely been employed in the tertiary sector, consistent with

the rapid growth of the tertiary sector as a share of the overall labour market (Baer 2008: chap.

18, World Bank 2002b).

[Figure 4 about here]

Second, we find that women tend to be more concentrated in a few occupations (e.g.

teaching associates, personal service workers), while non-whites are more homogenously

distributed in the occupational structure. Interestingly, we find that female dominated

occupations are generally more-skilled (such as teaching) than male dominated professions (such

as extraction and building trade workers). Along the same lines, non-white dominated

occupations are generally less skilled occupations than those that are white dominated. That

said, we also find that while women tend to be concentrated in particular occupations, those

                                                        16 Brito and de Carvalho (2006) and Gomes Braga (2006) explore the features of internal migration in Brazil, and report evidence of migration toward urban areas and toward the southern regions particularly during the 1990s. On the other hand, new work by Pochmann (2007) reports a different trend in recent years, as new regions, such as Amazonas, Mato Grosso e Goiás (among other, primarily in the Central-West and North regions) have replaced the South and the South-East regions as the primary recipients of internal migrants.

14  

specific occupations are quite different between the formal and informal sectors, with, for

example, large numbers of women in the informal employed as housekeepers or manufacturing

workers, while in the formal sector women are concentrated among teachers and clerks.

Third, we, not surprisingly, find some evidence that non-formal employment tends to be

concentrated in a relatively smaller number of occupations than formal sector employment. This

is suggestive of less diversified informal and self-employed sectors, but the difference with the

formal sector is less than we might have expected. The absence of a larger difference is in line

with the hypothesis, proposed by Bosch et al (2007), that informality does not simply exist in

marginal sectors, but frequently expands across all sectors of the labour market.

Finally, we find that the concentration of individual population groups within particular

occupations is declining over time, even within those occupations that remain highly segregated.

For example, female dominated occupations, such as personal services, have seen the increasing

entry of male workers, while financial services, which are highly male dominated, have

witnessed growing female participation. This trend towards greater homogeneity in the

representation of female and non-white labourers in the labour force is an important issue to

which we return later to the analysis of the driving forces behind changes in segregation over

time.

5. Measuring Occupational Segregation over Time

The analysis of the occupational segregation is undertaken by using three different

indices of segregation: the Duncan index (ID), the Karmel and Maclachlan index (IKM), the Gini

segregation index (IG). Our motivation in applying this wide range of measures is twofold: first,

to understand the extent to which the results may be dependent on the particular measures being

employed, and, second, to gain a corresponding insight into which measure is best suited to our

purpose.

The dissimilarity index, or Duncan index (Duncan and Duncan 1955), is certainly one of

the most applied indexes of segregation and is given by the following formula:

∑ with i=1,2,...,n (1)

15  

where Fi and Mi are the number of female and male workers in the ith occupation and F and B are

the total number of women and men in the labour force.17 The index is generally interpreted as

measuring the proportion of the female workforce that would be required to shift between

occupations in order to equalize female and male representation across occupations. The main

weakness is the fact that redistributing the female workforce to reach zero segregation there

would also imply a change in the occupational structure. Furthermore, this index assigns equal

weights to each occupation independent of its relative size. Watts (1998) claims that the Duncan

index fails to show occupation invariance, but it invariant to gender composition of the labour

force.

There are several other measures of occupational segregation that have been proposed in

the literature, of which we test two alternatives that address some of the criticisms of the Duncan

index. A modification of the dissimilarity index has been developed by Karmel and Maclachlan

(1988). The Karmel and Maclachlan index denotes the total labour force to be relocated with

replacement in order to reach zero female-male segregation, while keeping the occupational

structure and the overall female and male shares of the workforce constant.

∑ 1 (2)

where is the share of female in the total labour force.

Finally, the Gini segregation index is defined by the following formula:

∑ ∑

(3)

As Silber (1989a, 1989b) notes, the G-segregation index is equal to the Gini Index of the

female-male ratio, where the weights are the shares of each occupation in the total male

workforce. Assuming a segregation curve which can be defined as a cumulative distribution of

the proportion of women in every occupation, the index is given by twice the area lying between

the segregation curve and the equi-distribution line given by the 45 degrees diagonal in a similar

spirit to the Lorenz curve in the inequality literature.

                                                        17 The formulas reported in this section refer to gender segregation. In order to compute the indices for racial segregation F and Fi have to be intended for non-white workers and M and Mi with white ones.

16  

5.1 Occupational Segregation across Formal and Non-Formal Sectors

Turning to the results, each index is computed independently for both gender and racial

segregation, as well as separately for the formal, informal and self-employed sectors. Table 1

provides the results of computing the different measures of occupational segregation, along with

their bootstrapped standard errors.18 All indices of segregation have been computed using our

harmonized 3-digit occupational classification.19 In order to assess the importance of changes in

the segregation measures across sectors and over time, we calculate the statistical significance of

the changes among formal, informal and self-employed sectors in any given year as well as over

time, using five reference years (1987, 1992, 1997, 2002 and 2006). Tables A2 and A3 in the

appendix report the test of mean differences, respectively, among sectors and over time using the

standard parametric two sample t-test. Finally, in order to ensure that the results are robust, and

not driven by the specific methodological choices involved in constructing the new classification

of occupational codes, we conduct two robustness checks. The first compares the results to those

that emerge when employing the original PNAD classification, while the second checks the

impact of the exclusion of ‘zero wage’ observations from the analysis.20 The robustness checks,

though not reported in full here due to the constraints of space, confirm the credibility of the

results.

                                                        18 The standard errors are computed using the boostrapping technique based on 500 replications, following the approach illustrated by Efron and Tibshirani (1991, 1993). The bootstrap method estimates the distribution of the segregation measure by resampling with replacement in order to create multiple estimates of the statistics. These distributions are then used to construct confidence intervals around the original points and ultimately to establish standard errors (see also Boisso et al, 1994). 19 Figures on occupational segregation by using occupational codes at 2-digit level are also available: the patterns are almost the same, but the extent of segregation is on average smaller. The more detailed are the occupational categorization the greater is the outcome from any measures of segregation. 20 First, it is important to ensure that our findings represent actual changes in the distribution of workers across occupations, and are not an artefact of the methodology employed in constructing our new harmonized occupational classification. As a partial check against this possibility, our first robustness check compares our results presented in this section to results computed using the original classification: overall, the outcome of this comparison is reassuring. The second check refers to the exclusion of zero wage observations. The zero wage observations comprise missing wages and non-remunerated workers. Missing wages are on average only 1.4% of the entire sample and randomly distributed across occupations. Not remunerated workers represent instead an average of 9.5% of the sample. More importantly, not remunerated workers are non-random, and generally report employment in own-production, own-construction or as member of the household, mainly in the agricultural sector and are overwhelmingly women and primarily non-whites. In their earlier study of Brazilian informality, Ramos and Ferreira (2005) find that the exclusion of not remunerated workers from the analysis leads both to an underestimate of the size of the informal sector and to a change in the observed trend over time. Our findings are very similar, as we find that if we add not remunerated workers to the sample the overall estimate of the size of the informal sector increases as well as of occupational segregation - which is what we would expect given that the majority of not remunerated workers are female and non white.

17  

[Table 1 about here]

The results are, encouragingly, broadly similar across the three indexes,21 thus reinforcing

our confidence in the results. For simplicity, what follows focuses on the widely used Duncan

Index, and we can highlight three main findings. First, gender occupational segregation is always

considerably greater than racial occupational segregation - roughly three times greater. In 2006

the Duncan index between female and male workers was 0.565, which is much greater than the

Duncan index of 0.191 between non-whites and whites. This means that in 2006 more than half

of female workers and one fifth of non-white workers would have needed to be reallocated in

order to equalize workers’ representation across occupations.

Second, gender segregation is generally more severe in the informal and self-employed

sectors, as the Duncan index by gender in 2006 was 0.513 in the formal sector, while it was

0.653 for the informal sector. The same does not hold equally true for racial segregation, as

levels of segregation in the formal and informal sectors were not statistically different from each

other in 1987 and 2002, while during the 1990s racial segregation in the formal sector was, in

fact, slightly higher than in the informal sector. That said, this trend has been significantly

reversed since the beginning of the 2000s, when racial segregation in the informal sector began

to exceed that in the formal sector. These results are all statistically significant and hold among

each of the Duncan, Karmel & Machlachlan and Gini segregation indices (Table A2 in the

appendix).

Third, overall levels of segregation are declining, though this decline has been much

more pronounced for gender segregation, and in the formal sector rather than the informal sector.

Using the Duncan index, gender segregation decreases overall by 6.5% between 1987 and 2006,

while racial segregation declines by only 4.2%. Focusing on gender segregation, after a

negligible increase in the early 1990s, we record a decrease in all of the segregation measures

between the beginning of the 1990s and 2006, and these changes are always statistically

significant (Table A3). This decline in gender segregation is focused in the formal sector (-7.7%)

rather than in the informal sector (-5.5%). This declining trend for gender occupational

segregation in Brazil is consistent with Tzannatos (1999) who finds that in developing countries

                                                        21 Differences in segregation measures are mainly imputable to the properties of these indices and are acknowledged at the end of the section.

18  

gender differentials in employment and pay are narrowing even faster than was the case in

developed countries during the 1960 and 1970s, when they were experiencing rapid labour

market changes (Watts 1998, Baunauch 2002). Racial segregation similarly decreases primarily

in the formal sector (-8.2%) and is particularly sensitive to informality, as its decline in the

informal sector is small (-3.1%). In fact, in the informal sector we see a somewhat surprising,

and statistically significant, increase in racial segregation from the beginning of the 1990s,

resulting in an overall decline in racial segregation in the informal sector between 1987 and 2006

that is not statistically different from zero (Table A3). While Table 1 reports the measures of

occupational segregation every 5 years, Figure 5 plots the continuous evolution over time of the

different measures.

[Figure 5 about here]

Finally, although the trends described so far are broadly common across all of the

segregation indexes, it is also important to comment briefly on the differences in levels and

trends across them. As regard to comparison among segregation indices, the Gini segregation

index reports the highest figures among all indices. In 2006 the Gini segregation index was

equal to 0.735 and 0.262, by gender and race respectively. Instead of looking at mean deviations

as it occurs in the case of Duncan-type of indices, the Gini index uses mean differences to

measure the dispersion of the occupational distribution. Thus, segregation appears a more severe

problem when focusing on the compositional differences among all occupations, as the Gini

index does, rather than focusing on how gender and racial ratio varies between each occupation

and the overall workforce composition, as in the Duncan index.

By using the Karmel and Maclachlan index, the level of estimated segregation

dramatically diminishes although the patterns are in line with the other indices. Again, the

reason can be found in how these indices are constructed. The Duncan index detects the number

of female workers that need to be moved without replacement allowing changes in the

occupational distribution. The Karmel and Maclachlan index measures the number of female

workers that should be shifted with replacement in order to obtain zero segregation without

changing the relative size of each occupation and the overall size of the labour force (Watts,

1998). In Table 1, we notice that when the Duncan index is decreasing, the Karmel and

19  

Maclachlan index decreases less or in some cases it even increases. This implies that, although

the female-male differential has narrowed, the increasing number of women entering the labour

market has meant that the proportion of workers needed to shift occupations to reach zero

segregation has not changed or it has slightly increased (Karmel and Maclachlan, 1988).

5.2 Exploring Demographic, Educational, Sectoral and Spatial Patterns

As with the previous section, we conclude by exploring trends in occupations segregation

disaggregated by key characteristics of the population (demographic, educational, sectoral and

spatial). We calculate the Duncan index over time for each of the sub-groups of interest in order

to disentangle differential trends in the data. The findings are presented in Figure 6 and we

simply summarized the most important findings here.

In terms of demographic patterns, gender segregation is higher among older workers,

while racial segregation is, somewhat surprisingly, higher among younger workers. With respect

to educational patterns, gender segregation is lower among those with more education, but, racial

segregation is, surprisingly higher among those with more education. This finding suggests that

not only are non-whites a comparatively small share of the highly educated workforce, but they

are unusually concentrated in some professions, and absent from others. That said, while

segregation is higher among highly educated non-whites, we also see that after increasing rapidly

from 1987-1993, segregation within this group has declined rapidly since the mid 1990s, while

segregation among less educated groups has been stable, or slightly increasing, over time. We

observe a similar trend among women, as gender segregation is not only lower among highly

educated workers but it has also been declining rapidly, whereas segregation has declined only

very modestly for less educated groups.

Moving to sectoral patterns, we find that segregation has increased in the secondary

sector, particularly with respect to gender, while the opposite is true in the primary sector, where

gender segregation has declined significantly. In the tertiary sector both gender and racial

segregations have consistently declined over time. Within the tertiary sector women and non-

whites have continued to make up the majority of the labour force, even as the overall size of the

tertiary sector has grown rapidly, reflecting rapid new entry in this sector. The fact that

segregation has been declining thus seems to reflect decreasing concentration within individual

20  

occupations, as white men in particular also enter the sector and join traditionally female and

non-white dominated professions.22

Turning finally to spatial patterns of segregation, we see that segregation has declined

relatively homogenously across all regions. That said, there has been a particularly dramatic

decline in gender segregation in the Central-West region, while racial segregation has been

declining everywhere, but remains strikingly much higher in the South-East and South regions

than elsewhere in the country. Finally, gender segregation is higher in rural areas while racial

segregation is higher in urban areas.

[Figure 6 about here]

6. Decomposition of changes in segregation over time

Having presented these broad trends, what remains is to seek to better understand the

underlying forces that have driven these changes in occupational segregation over time. In

particular, the analysis to follow seeks to understand whether declining occupational segregation

is driven by more homogenous composition by gender and race within individual occupations,

by changes in occupational structure (namely occupations’ weights) or by changes in sub-

population shares (gender or racial) of the entire labour force. This is particularly crucial given

that the Duncan index, as well as other indices of segregation, is sensitive to changes in

occupations’ weights and to changes in gender and racial shares of the labour force. We thus

employ the decomposition methodology proposed by Deutsch et al (2009), which combines the

Karmel and Maclachlan (1988) decomposition and the concept of the Shapley value in order to

distinguish between these three different sources of variation in occupational segregation (see

also Shorrocks, 1999 and Sastre and Trannoy, 2002).

Segregation can change over time because of the changes in the internal gender or racial

composition within each occupation. This source of variation is also termed ‘net segregation’ or

variation in the ‘internal structure’ because it is independent of the variations that can occur ‘in

                                                        22 The tertiary sector covered 43% of the entire economy in 1987 and had grown to 67% in 2006. The female share of the tertiary sector has declined from 56% in 1987 to 52% in 2006, while the non-white share has increased, moving from 46.4% to 51%.

21  

the margins’. The changes in segregation occurring because of ‘variation in the margins’ capture,

respectively, changes in the relative weights of occupations or changes in the shares of sub-

populations groups in the total labour force. The sum of these three sources of variation (i.e. the

internal structure and the two components of the margins) is termed ‘gross variation’ in

occupational segregation.23

Following Deutsch et al (2009) derivation, it is possible to decompose the change over

time of a segregation index as follows:

∆ (4)

where and represent, respectively, the indices for the final and initial periods of time. If we

apply the concept of the Shapley decomposition following Deutsch et al (2009), the total

variation, defining ‘gross variation’ of segregation over time, can be decomposed as follows:

∆ ∆ , ∆ ∆ ∆ (5)

where ∆ and ∆ represent the two main components of the decomposition, the component of

the change due to the variation in the ‘margins’ and the component of the change due to

variations in the ‘internal structure’ (or ‘net segregation’) and they will be

∆ ∆ ∆ , ∆ ∆ (6)

and

∆ ∆ ∆ , ∆ ∆ (7)

The contribution of these components is then proved to be expressed also as follows:

∆ (8)

and

∆ (9)

where the set of matrices employed in the above equations are obtained by interacting both

margins and internal structure of the segregation matrices from which the two indices and

can be drawn. The two initial matrices are P and V and we need to compare them to derive

matrix S which has the internal structure of P but the margins of V. In the same way, matrix W

                                                        23 The Shapley decomposition by Deutsch et al (2009) is inspired by the decomposition technique proposed in Karmel and Maclachlan (1988). In fact, they have proposed to decompose the segregation index into the mix effects (gender, occupation and gender by occupation) and the composition effect, which are similar respectively to the variations due to the margins and to the internal structure proposed by Deutsch et al (2009). The important innovation of the Shapley decomposition is the absence of an interaction term or residual from the decomposition.

22  

can be derived with the internal structure of matrix V and the margins of matrix P, simply by

inverting the process.

In order to explain the derivation, let’s start by considering the matrix P. This matrix has

the ratio in its internal structure where is the number of individuals of occupation i and

from the subpopulation j and T is the total number of workers. The margins of matrix P are

defined by · and · which are respectively the horizontal margins

(occupational structure) and the vertical margins (shares of the sub-populations).

To derive the matrix S, we need to multiply all element of P by the ratio ·· and obtain

an intermediate matrix X. Its elements need to be multiplied by the ratio ··

to obtain a new

matrix Y and so on, after several iterations, the matrix will converge to the matrix S with the

internal structure of P and the margins of V (see Deming and Stephan, 1940).

As said, we could also start with the matrix V and by applying the same procedure we will end

up with the matrix W that has the internal structure of matrix V and the margins of matrix P.

Now, the proposed decomposition is then able to decompose the variation in the margins

into the component due to the variation in the occupational structure and the shares of the sub-

populations. In other words, we will have that

∆ ∆ ∆ (10)

where represent the contribution from changes in occupational structure and from changes in

shares of sub-populations. Using the same procedure as before we can express these two

components as follows:

∆ (11)

and

∆ (12)

In order to derive these components we need to define additional matrices (see Deutsch et al

(2009) for the detailed construction of these matrices):

matrix L with the internal structure of P, the horizontal margins of V and the vertical

margins of P;

23  

matrix K with the internal structure of P, the horizontal margins of P and the vertical

margins of V;

matrix F with the internal structure of V, the horizontal margins of V and the vertical

margins of P;

matrix C with the internal structure of V, the horizontal margins of P and the vertical

margins of V.

Through this decomposition, we are hence able to decompose the change between two

periods into

∆ ∆ ∆ ∆ (13)

the variation due to the sub-population shares within occupations (the net segregation or changes

in internal structure), ∆ , the variation due to the occupational structure of the labour markets

(i.e. the weights of each occupation), ∆ , and finally the sub-population shares of the total

labour force (i.e. gender or racial composition of the labour force), ∆ .

6.1 Empirical Findings across Formal and Non-Formal labour markets

We perform the decomposition of changes in gender and racial segregation using the

Duncan index between two periods: the initial period, comprising the years 1987, 1988, 1989,

1990 and 1992, and the final period, comprising the years 2002 to 2006. We aggregate the first

and last five years in order to have a sufficient number of observations to implement the

decomposition separately across the formal, informal and self-employed labour markets, as well

as disaggregated by key characteristics of the labour force.24 This aggregation does not appear to

be problematic, as changes in occupational distribution within the aggregated years are relatively

modest. Finally, we also compute bootstrapped standard errors for the overall changes in

occupational segregation, as well as for the components of these changes, based on 500 random

samples, in order to test the statistical significance of each component. The findings are reported

in Table 2, and in Figure 7.

[Table 2 about here]

                                                        24 For the sake of brevity, we do not report the Shapley decomposition results for the Karmel and Maclachlan index and the Gini segregation index, as the results are largely unchanged, when statistically significant, relative to the results presented here (they are available from the author).

24  

In general, we observe that the decline in both gender and racial segregation, which is

also called the ‘gross variation’, is driven overwhelmingly by ‘variations in the internal

structure’, also called ‘net variation’ in segregation – that is, declining concentration by gender

and race within individual occupations. The contribution of the internal structure component is

almost always statistically significant as visible in Table 2. The fact that declining segregation is

driven by improvements in the gender and racial composition within occupations is consistent

with the arguments made in at least a few similar studies in the developed world (see for

example Blau and Hendricks (1979) and more recently Queneau (2009)). By contrast, we find

that the impact of ‘variations in the margins’ (that is, changes in occupational structure -

occupations’ weights - and in the share of different population sub-groups in the overall labour

force) is generally to increase levels of occupational segregation. However, the ‘variations in the

margins’ component warrants more careful analysis, as its two components behave differently

across the formal and non-formal labour markets. We look first at the formal, informal and self-

employed sectors separately, and then consider the labour market as a whole.

In the formal sector, when looking at both gender and racial segregation, changes in

occupational structure (occupations’ weights), have contributed to declining segregation,

although the effect is only statistically significant in the case of gender. On the other hand, the

non-formal sectors show the opposite trend, as changes in occupations’ weights are the main

cause of upward pressure on levels of both racial and gender segregation, with a particularly

dramatic effect in the case of racial segregation. This trend is particularly pronounced for the

self-employed sector, where the increase in segregation caused by changes in occupations’

weights completely offsets the decline in occupational segregation resulting from variations in

the internal structure of occupations, leading to an increase in racial segregation for self-

employed workers. All of these findings in the non-formal sector are statistically significant.

The last component, changes in the sub-population shares, generally contributes to

increased segregation, with the exception of gender segregation in the informal sector, where the

increase in female participation positively contributes to reducing gender segregation (the same

happens in the self-employed sector, but the component is not statistically significant in that

case). That said, this component is consistently comparatively small in magnitude.

When we combine the formal, informal and self-employed sector and look at the labour

market as a whole, we see that the aggregate affect of the changes in occupational structure

25  

component differs between gender and racial segregation. Focusing first on gender segregation,

we see that the aggregate effect of changes in occupational structure is to reduce segregation, as

the trend in the formal sector (reducing segregation) outweighs the trend that we observe in the

informal sector (increasing segregation). This is consistent with the fact that for women

experience a larger increase in participation in the formal sector (from 34% to almost 43%) than

in the informal sector (from 42.2% to 47.5%), while the formal sector is less segregated than the

informal sector, and, as such, the growth of the formal sector is likely to result in a less

segregated occupational structure overall. Turning to racial segregation, the results are more

straightforward with: variation in the internal structure reduces segregation, as is the case in each

sector on its own, while the ‘variation in the margins’ component increases segregation,

consistent with the fact that this component increases segregation in the non-formal labour

markets and is statistically insignificant in the formal sector.

[Figure 7 about here]

Having laid out these broad findings about the determinants of gender and racial

segregation, it remains to support the findings of the decomposition exercise with reference to

the descriptive data on occupational structure and segregation presented earlier.

The most important finding is that the primary driver of falling occupational segregation

is variations in internal structure, and this is consistent with the descriptive data explored earlier.

Looking first at gender segregation, almost all of the most female dominated occupations in the

labour market have experienced a decreasing female share over time, as men have increasingly

entered these professions. For example, 93.45% of teaching associate professionals were women

in 1987, while this share had fallen to 82.8% in 2006; customer services clerks moved from a

female share equal of 83.13% in 1987 to 75.19% in 2006. By contrast, while a small number of

male dominated occupations have remained almost closed to women (e.g. drivers and mobile

plant operators, extraction and building trades workers, metal and machinery related trades

workers), other male dominated occupations have seen a significant entry of female workers (e.g.

physics, engineers and sales persons).

Turning to racial segregation, variations in internal structure are equally important,

though the sources of this variation are slightly different. Among occupations historically

26  

dominated by non-whites, the share of non-white labourers has declined in some areas (for

example, in mining, construction, manufacturing and transports the non-white share fell from

80.36% in 1987, to only 59.93% in 2006), but other non-white dominated professions have seen

little change in their composition over time, in part because the extent of segregation in these

occupations is comparatively low, as non-white workers heavily dominate very few occupations

in Brazil. As such, it appears that another important source of declining within occupation

segregation has been the growing share of non-whites in white dominated professions, as this is

common across almost all of the occupations that have historically been dominated by white

workers (e.g. life science and health professionals and teachers).

While variations in internal structure have driven declining segregation, we find that

changes in the margins have, on average, increased occupational segregation. Most importantly,

across both gender and racial segregation, changes in occupations’ weights have contributed to

increasing levels of segregation, though this effect is concentrated entirely in the non-formal

labour markets. The implication is that within informal labour markets relatively segregated

occupations have grown larger over time, thus increasing occupational segregation, though this

has not been the case in the formal sector. Looking at gender segregation we find that the largest

and most female dominated occupations (namely cod. 512, 514, 522) have grown rapidly,

particularly as non-formal activities, led by housekeepers and restaurant workers (cod. 512) (see

Figure 4). We find similar patterns in terms of racial segregation, where the overall impact of

variations in occupations’ weights in increasing segregation is relatively greater. As with gender

segregation, we see that the most rapidly growing occupations, housekeepers and restaurant

workers (cod.512) and non-self-employed agricultural occupations (cod. 612), have high and

increasing shares of non-white workers and have had their growth concentrated in the informal

labour market. Thus, in aggregate we see that the growth of relatively segregated occupations in

the tertiary sector has contributed to increasing segregation, but this has been more than offset by

greater equity in the internal composition of these occupations over time (for a similar finding

see Tomaskovic-Devey et al (2006) who look at the role of the service economy in reducing

gender segregation in the US).

Finally, the data suggests, somewhat counter-intuitively, that, after accounting for the

other trends discussed so far, the increasing share of women and non-whites in the labour force

has contributed to increasing segregation, implying that new entrants to the labour force have

27  

disproportionately entered occupations in which women and non-whites, respectively, were

already dominant. This is revealed, for example, in the continued entry of women, and

particularly non-white women, into the already dominated housekeeping profession. That said,

the magnitude of these effects is comparatively very small.

6.2 Empirical Findings Disaggregated by Characteristics of the Labour Force

Finally, as with the earlier analysis of occupational segregation, we decompose changes

in gender and racial segregation disaggregated by several key characteristics of the labour force.

Table 3 reports the estimated components of the decomposition together with their boostrapped

standard errors. As many of these results are not statistically significant, we comment briefly

only on the most relevant results.

When we divide the labour force by age groups we find that changes in internal structure

consistently reduce segregation across all groups. Turning to the margins, it is only among the

elderly that changes in occupational structure have had a statistically significant impact on

occupation segregation, pushing it upwards. By contrast, in the case of racial segregation,

changes in the occupational structure increase racial segregation for all age groups, while the

negative impact is relatively stronger among young people that is offsets the positive effect of

improvements in the internal structure of occupations, leading to an overall increase in labour

segregation among young people.

Looking at educational attainment, and considering only the statistically significant

components, we find that among more educated workers, gender segregation decreases because

of the positive contribution of changes in internal structure, while the negative contribution from

changes in occupational structure is negligible. Conversely, racial segregation increases among

the well educated because the contribution of changes in internal structure is small and

completely offset by the huge negative contribution of changes in occupational structure. This

again captured the dominance of white workers in more educated and skilled occupations, while

also pointing towards the growth of such white dominated professions over time (for example,

financial services).

When looking at sectoral patterns, we find that changes in occupational structure have

contributed to increasing gender segregation only in the secondary sector. Consistent with this

view, fast growing occupations in the secondary sectors have included, for example, food

28  

processing, wood treaters, textile processing related jobs as well as machine and plant operators

and assemblers, all of which are highly gender segregated. In terms of racial segregation, the

only statistically significant decomposition finding is within the tertiary sector, where changes in

internal structure have strikingly reduced segregation.

When we turn to spatial patterns, and separate rural and urban areas, we notice that

gender segregation has decreased in both rural and urban areas, driven by the positive

contribution of changes in internal structure, while, particularly in urban areas, the negative

contribution of changes in occupational structure is negligible. With respect to racial

segregation, we record a consistent decrease in urban areas, driven by a clear improvement of the

internal structure within occupations.

[Table 3 about here]

7. Final Discussions and Conclusions

This study has yielded a wide range of new insights about occupational segregation in

Brazil. This reflects both the use of a more complete, and more accurate, dataset, and the

presentation of a more nuanced and disaggregated view of occupational segregation over time.

This section briefly summarizes certain key trends, while pulling together these disparate

findings in order to highlight a number of important conclusions.

The results highlight several major trends in the labour market that provide the

background to our conclusions about occupational segregation. First, we find a large increase in

female participation, which is common for Latin American countries where the gap between

male and female participation has narrowed more than in any other region in the developed

world (Wajman and Rios Neto 2000, Soares and Izaki 2002). Second, we find a similarly rapid

increase in the share of non-white workers, who have comprised the majority of the work force

since 2003.25 Third, we find that the non-formal labour markets have remained relatively

                                                        25 Whether the number of non-white individuals entering the labour force has really increased over time or whether the number of individuals among work force that report themselves as non-white population is increasing cannot be traced, as in the PNAD dataset the race/skin colour is self-reported. This finding hence should be interpreted with caution. 

29  

constant as a share of the total labour force between 1987 and 2006. This contrasts with some

studies that report rising informality, but, like Ramos and Ferreira (2005), we find that this is a

largely a question of sample selection, as Brazilian informality is concentrated in metropolitan

areas. Fourth, we observe a very rapid growth of the tertiary sector, which absorbs almost all of

the new female and non-white entrants into the labour force, thus leaving the occupational

distribution surprisingly stable despite major changes in the composition of the labour force

(Baer 2008: chap.18, World Bank 2002a, 2002b).

Finally, we find that the share of women, and particularly non-white women, in the

informal sector has increased significantly over time. It may be that key features of informal

sector employment, such as flexibility, lower commitment to long-term job positions and higher

turnover, are well suited to female labour supply in terms of preferences and tastes. On the other

hand, it may be that there are high barriers to entry into the formal sector for women, and that the

informal market might correspondingly exploit the lack of choice available to less skilled female

workers, mainly non-white, employed in personal services, such as housekeepers.26 This in

consistent with Telles’ (1992) earlier finding that “[...] education and race are more frequently

used in screening women’s than men’s entrance into the formal sector”.

It is against this background, that the paper highlights a range of novel findings about the

extent, evolution and characteristics of occupational segregation. In broad terms, we find that

gender segregation is much larger, in absolute terms, than racial segregation, while both have

declined over time. This initially seems to point towards particularly unequal opportunities for

women, but it must also be borne in mind that an important part of this segregation may be

explained by differences in tastes and preferences between men and women (see also the

discussion in Bertrand, 2010). By contrast, while segregation by race is significantly lower in

absolute terms, it may actually be a more serious problem, as it cannot be as easily explained by

differences in tastes and preferences, and has been comparatively persistent over time despite the

rapid entry of non-whites into the labour force.

                                                        26 The growing role of non-white women in the personal services sector, as housekeepers and often in the informal sector, has been widely noted in this paper. While not central to the evidence presented here, it is worth noting two additional trends in this area that have gained recent attention. First, as noted earlier, despite the high level of informality attached to this type of occupation, the ILO reports that we are witnessing a significant trend of ‘formalization’ in recent years (Berg, 2010; ILO, 2010). Second, it has been noted that the income of domestic employees rose 34 percent from 2003 to 2009, which is more than twice the average increase for all of Brazil’s active workers, while their working hours fell by 5 percent to 36.2 hours a week. (see New York Times article: http://www.nytimes.com/2011/05/20/world/americas/20brazil.html?_r=1)

30  

We gain deeper insight into the particular characteristics of this occupational segregation

when we disaggregate the findings into the formal, informal and self-employed sectors, and by

key characteristics of the population. Looking first at gender segregation, the patterns are

relatively consistent with expectations: segregation has declined in both the formal and informal

sectors, and has declined more among young workers and among the better educated. That is,

new opportunities are most available to younger, educated, female workers, but there has also,

encouragingly, been an important decline in segregation among other workers as well. We

observe that gender segregation has widely declined in both the primary and tertiary sectors, but

it has been stubbornly high, and modestly increasing, in the secondary sector.

When we turn to racial segregation the trends are much more mixed, reinforcing the

notion that, while smaller in magnitude, racial segregation may pose a particular challenge.

First, while racial segregation has declined in the formal sector, it has experienced only a

negligible decline overall, while it has been increasing in the informal sector in recent years.

Second, whereas women are well represented in, and even dominate, many highly skilled

occupations, non-whites are heavily concentrated in lower skill occupations and racial

segregation is, surprisingly, higher among the better educated. That said, while racial

segregation increased rapidly among the highly educated from 1987-1993, it has been declining

rapidly since then. Third, and even more surprisingly, we find that racial segregation is higher

among younger workers, which appears to be an even starker indicator of the relative persistence

of racial segregation, and the apparent barriers faced by less experienced non-white workers.

Finally, we find that racial segregation is higher in urban areas and in the South and South-East

regions, suggesting that it may be non-white migrants who are particularly concentrated in

particular professions.

Having described these trends in segregation, the most novel results emerge from our

application of the Shapley decomposition, as proposed by Deutsch et al (2009), to identify the

forces driving declining segregation. Our results indicate very clearly that the decline in both

gender and racial segregation is overwhelmingly the consequences of more homogenous gender

and racial composition within occupations. On the other hand, we also find that, particularly in

the non-formal labour markets, these improvements are partially offset by changes in

occupational structure and the entry of new groups into the labour force, both of which have

contributed to increasing segregation, with many new entrants to the labour force joining

31  

traditionally more segregated occupations, which have significantly increased in size over time.

Our aggregate results are driven by the fact that the increase in segregation provoked by these

two latter trends is offset by the general improvement in composition within each occupation.

This in many ways represents a more ‘real’ decline in segregation, and is thus a very

encouraging finding from a social perspective.

Finally, it is useful to conclude with a forward looking note about what these results

suggest about the possible impact of anti-discrimination legislation (ADL). As was noted at the

outset, this is a very difficult question to study directly, and has, consequently, been the subject

of limited empirical research in developing countries.27 The results presented here nonetheless

provide important initial insights, resulting primarily from disaggregating the formal and non-

formal sectors, as it is only in the former that we expect ADL to play a role. And, indeed, our

broad findings are consistent with the view that ADL has, indeed, played a role.

Most obviously, occupational segregation, by both gender and race, is declining more

rapidly in the formal sector than in the non-formal sectors. In this view, and following the

results of the Shapley decomposition, ADL has not only contributed to significant improvements

in the internal structure of individual occupations, but has also restricted the emergence and

growth of highly segregated occupations within informal activities. Thus, for example, we see

that among housekeepers and restaurant workers, which are female and non-white dominated

occupations, there is always significantly less segregation in the formal sector than in the

informal sector. Finally, less concretely, but certainly provocatively, we also observe a dramatic

decline in gender segregation in the Central West province, and particularly in the Distrito

Federal, where we might expect the impact of government policy, and thus ADL, to be

strongest. In addition, this decline might have something to do with the fact that the government

is the major employer in the Distrito Federal, and to get a job in the civil service one has to go

through examinations which are race and gender blind (Osorio, 2006).

This said, the results presented here are also open to an alternative interpretation, which is

also consistent with the Shapley results, and this is that ADL, and government regulation more

generally might not have played an important role in reducing segregation, but may primarily

have led more segregated occupations to function in the informal sector, where they have grown

                                                        27 Interesting studies applied to developed countries are, among others, Heckman and Payner (1989) and Neumark and Stock (2006).

32  

rapidly. In this view, which is consistent with research noted earlier on the impact of regulation

of levels of informality, while ADL may reduce segregation when it is enforced, in practice the

real impact may be small, as segregated occupations may simply choose to function in the

informal sector, thus explaining the rapid growth of highly segregated occupations in that sector.

While this paper has thus presented the most significant and detailed evidence to date on the

connections between formality, informality and occupational segregation, it remains an area rich

with opportunities for further research.

33  

References

Abramo, L. (2004) “Desigualdades e discriminacao de genero e raca no Mercado de

trabalho brasileiro e suas implicacoes para a formulacao de uma politica de emprego” Testo

elaborado para o Seminario Nacional: Politica peral de emprego: Necesidades, opcoes,

prioridades, OIT, Brasilia, 9 e 10 de dezembro de 2004

Albelda, R. (1986) “Occupational segregation by race and gender, 1958-1981” Industrial

and Labor Relations Review 39(3): 404-411

Almeida, R., and Carneiro P. (2007) “Inequality and Employment in a Dual Economy:

Enforcement of Labor Regulation in Brazil” IZA Discussion Paper No. 3094

Anker, R. (1997) “Theories of occupational segregation by sex: An overview”

International Labour Review 136(3)

Anker, R., Melkas H., and Korten A. (2003) “Gender-based occupational segregation in

the 1990’s” Declaration Working Paper No.16. Geneva, ILO

Arabsheibani, G. R., Carneiro F. G., and Henley A. (2003) “Gender Wage Differentials in

Brazil: Trends over a Turbulent Era” World Bank Policy Research Working Paper 3148

Arcand, J.L., and D’Hombres B. (2004) “Racial Discrimination in the Brazilian Labour

Market: Wage, Employment and Segregation Effects” Journal of International Development

16:1053-1066

Arias, O., G. Yamada, and Tejerina L. (2004) “Education, family background and racial

earnings inequality in Brazil” International Journal of Manpower 25(3/4): 355-374

Baer, W. (2008) “The Brazilian Economy: Growth and Development” Westport,

Connecticut: Praeger

Barros, R.P., Machado A.F., Mendonca, R.S.P. (1997) “A desigualdade da pobreza:

estrategias ocupacionais e diferenciais por genero” IPEA Texto para discussão No. 453

Baunach, D. M. (2002) “Trends in Occupational Sex Segregation and Inequality, 1950 to

1990” Social Science Research 31: 77-98

Berg (2010) “Laws or luck? Understanding rising formality in Brazil in the 2000s” ILO

Working Paper No. 5, Serie Trabalho Decente no Brazil

Bertrand, M. (2010) “New Perspectives on Gender” Handbook of Labor Economics,

Volume 4b, Chapter 17

34  

Blackburn, R. M., Brooks B., and Jarman J. (2001) “The Vertical Dimension of

Occupational Segregation” Work Employment Society 15; 511-538

Blackburn, R. M., Jarman J. (2005) “Segregation and inequality” GeNet WP No.3

Blackburn, R. M., Jarman J., and Siltanen J. (1993) “The Analysis of Occupational

Gender Segregation over Time and Place: Considerations of Measurement and Some New

Evidence” Work, Employment and Society 7

Blau, F., D. and Hendricks, E. W. (1979) “Occupational Segregation by Sex: Trends and

Prospects” Journal of Human Resources 14(2): 197-210

Boisso, D., Hayes K., Hirschberg J., and Silber, J. (1994) “Occupational Segregation in

the Multidimensional Case: Decomposition and Tests of Statistical Significance” Journal of

Econometrics 61: 161-171

Bosch, M., Goni E., and Maloney, W. (2007), “The Determinants of Rising Informality in

Brazil: Evidence from Gross Worker Flows” World Bank Policy Research Working Paper No.

4375

Brito F., and de Carvalho, J. A. (2006) “As Migracoções Internas no Brasil: As

Novidades Sugeridas pelos Censos Demográficos de 1991 e pela PNADs Recentes” UFMG,

mimeo

Cacciamali, M. C. (1982) “Um estudo sobre o setor informal urbano e formas de

participacao na producao” USP, PhD thesis

Cacciamali, M. C. (1983) “Setor Informal Urbano e Formas de Participaçáo” Serie

Ensaios Economicos N. 20, IPE/USP

Cacciamali, M. C. (1988) “Mudanças Estruturais no Produto e Emprego no Brasil: 1950–

80” FEA/USP, mimeo

Calonico, S. and Ñopo, H. (2007), “Occupational Segregation by Gender and Wage Gaps.

Evidence from Urban Mexico 1987-2004”, IADB Working Paper 636

Carneiro, F. (1997) “The Changing Informal Labour Market in Brazil: Cyclicality versus

Excessive Intervention” Labour 11(1): 3-22

Carvalho, A., Néri M., Silva, D. (2006) “Diferenciais de Salários por Raça e Gênero no

Brasil: Aplicação dos Procedimentos de Oaxaca e Heckman em Pesquisas Amostrais

Complexas”, IBGE, mimeo.

35  

Castro, J. and Reilly B. (2010) “Occupational Segregation by Gender: An Empirical

Analysis for Urban Colombia (1986-2004)” University of Sussex, mimeo

Charles, M., and Grusky, D. B. (1995) “Models for Describing the Underlying Structure

of Sex Segregation” The American Journal of Sociology 100(4):931-971

Cotter, D. A., Hermsen J. M. and Vannerman, R. (2003) “The effects of occupational

gender segregation across Race” The Sociological Quartely 44(1): 17-36

Deutsch, J. and Silber, J. (2005) “Comparing segregation by gender in the labour force

across ten European countries in the late 1990s: An analysis based on the use of normative

segregation indices” International Journal of Manpower 26(3): 237-264

Deutsch, J., Fluckinger Y., and Silber, J. (1994) “Measuring occupational segregation.

Summary statistics and the impact of classification errors and aggregation” Journal of

Econometrics 61: 133-146

Deutsch, J., Fluckinger Y., and Silber, J. (2009) “Analysing Changes in Occupational

Segregation: the case of Switzerland 1970-2000”, in Occupational and Residential Segregation

Research on Economic Inequality, Volume 17, 173–204

Deutsch, R., Morrison, A., Piras, C. and Ñopo, H. (2005) “Working Within Confines:

Occupational Segregation by Gender in Three Latin American Countries” The Icfai Journal of

Applied Economics, IV(3): 50-59

Duncan, O., and Duncan, B. (1955) “A methodological analysis of segregation indexes”

American Sociological Review 20: 210-17

Efron, B., Tibshirani, R. (1991) “Statistical data analysis in the computer age” Science

253: 390-395

Efron, B., Tibshirani, R., (1993) “An Introduction to the Bootstrap” New York: Chapman

& Hall

Fiess, N. M., Fugazza, M. and Maloney, W. F. (2008) “Informality and Macroeconomic

Fluctuations”, IZA Working Paper No. 3519

Flückiger, Y., and Silber J. (1999) “The Measurement of Segregation in the Labor Force”

Heidelberg, New York: Physica-Verlag

Fonseca, M., and Rayp, G. (2011) “A view of trade liberalization on the agricultural labor

market in Brazil” Ghent University, mimeo

36  

Fryer, R. G. Jr. (2010) “The importance of segregation, discrimination, peer dynamics,

and identity in explaning trends in the racial achievement gap” NBER, Working Paper 16257

Gasparini, L., and Tornarolli, L. (2007) “Labor Informality in Latin America and the

Caribbean: Pattern and Trends from Household Survey Microdata” CEDLAS, mimeo

Goldberg, P. K., and Pavcnik, N. (2003) “The Response of the Informal Sector to Trade

Liberalization” Journal of Development Economics 72(2): 463-496

Gomes Braga F. (2006) “Migração Interna e Urbanização no Brasil Contemporâneo: Um

estudo da Rede de Localidades Centrais do Brasil (1980/2000)” Universidade Salgado de

Oliveira, mimeo

Hakim, C. (1992) “Explaining Trends in Occupational Segregation: The Measurement,

Causes, and Consequences of the Sexual Division of Labour” European Sociological Review,

8(2): 127-152

Hakim, C. (1993) “Segregated and Integrated Occupations: A new approach to Analysing

Social Change” European Sociological Review 9(3): 289-314

Hart, K. (1971) “Informal Income Opportunities and Urban Employment in Ghana” The

Journal of Modern African Studies 11(1): 61-89

Hart, K. (1972) “Employment, income and inequality: A strategy for increasing

productive employment in Kenya” ILO Report

Heckman, J. and Payner, B. (1989) “Determining the Impact of Federal

Antidiscrimination Policy on the Economic Status of Blacks: A Study of South Carolina” The

American Economic Review, 79(1): 138-177

Henley, A., Arabheibani, R. G., and Carneiro, F. (2009) “On Defining and Measuring the

Informal Sector: Evidence from Brazil” World Development 37(5): 992-1003

Hoek, J. (2002) “Labor Market Institutions and Restructuring: Evidence from Regulated

and Unregulated Labor Market in Brazil” William Davidson Working Paper Number 484

Hutchens, R. (1991) “Segregation Curves, Lorenz Curves, and Inequality in the

Distribution of People across Occupations” Mathematical Social Sciences 21:31-51

Hutchens, R. (2004) “On Measure of Segregation” International Economic Review 45(2)

ILO (2010) “Trabalho Domestico no Brasil” Brasilia, ILO Report

Karmel, T., and Maclachlan, M. (1988) “Occupational Sex Segregation – Increasing or

Decreasing?” The Economic Record 64: 187-95

37  

King, M. C. (1992) “Occupation segregation by race and sex, 1940-1988”, Monthly

Labor Review, U.S. Department of Labor, Bureau of Labor Statistics, 115(4): 30-37

Lago, L. (2006) “Trabalho, moradia e (i)mobilidade especial na metropole do Rio de

Janeiro” unpublished paper written for the XVI World Congress of the International Sociological

Association

Lewis, D. (1982) “The measurement of the Occupational and Industrial Segregation of

Women” Journal of Industrial Relations 24(3): 406-423

Lovell, P. (1994) “Race, Gender and Development in Brazil” Latin American Research

Review 29(3): 1-35

Lovell, P. (2000) “Race, Gender and Regional Labour Market Inequalities in Brazil”

Review of Social Economy 58(3): 277-293

Lovell, P. (2006) “Race, Gender, and Work in Sao Paolo, Brazil, 1960-2000” Latin

American Research Review 41(3)

Machado, A. F., De Olivera, A. and Carvalho, N. (2003) “Tipologia de Qualificação da

Força de Trabalho: uma Proposta a Partir da Noção de Impompatibilidade entre Ocupação e

Escolaridade” Texto para Discussão, 218, Universidade Federal de Minais Gerais

Maloney, W. F. (1999) “Does Informality Imply Segmentation in Urban Labor Markets?

Evidence from Sectoral Transitions in Mexico” The World Bank Economic Review 13(2): 275-

302

Maloney, W. F. (2004) “Informality Revisited” World Development 32(7): 1159-1178

Maume, D. J. (1999) “Glass Ceilings and Glass Escalators” Work and Occupations 26(4):

483-509

Melkas, H., and Anker, R. (1997) “Occupational segregation by sex in Nordic countries:

An empirical investigation” International Labour Review 136-341

Moir, H. and Selby Smith, J. (1979) “Industrial Segregation in the Australian Labour

Market” Journal of Industrial Relations 21(3): 281-291

Mora, R., and Ruiz-Castillo, J. (2003) “Additively decomposable segregation indexes.

The case of gender segregation by occupations and human capital levels in Spain” Journal of

Economic Inequality 1: 147–179

Muedler, M.-A., Poole, J., Ramey, G., and Wajnberg, T. (2004) “Job Concordances for

Brazil: Mapping the Classificacao Brasileira de Ocupacoes (CBO) to the International Standard

38  

Classification of Occupations (ISCO-88)” University of California, San Diego and CESifo,

mimeo

Neuman, S. (1994) “Ethnic Occupational Segregation in Israel” in S. Neuman and J.

Silber, eds., Inequality in Labor Markets: The Economics of Labor Market Segregation and

Discrimination, Research on Economic Inequality, Greenwich, CT and London, England: JAI

Press

Neuman, S. (1998) “Occupational segregation in the Israeli labour market: The gender-

ethnicity interaction” International Journal of Manpower 19(8)

Neumark, D., and Stock W. A. (2006) “The Labor Market Effects of Sex and Race

Discrimination Laws” Economic Inquiry 44(3): 385-419

Oliveira, A. M. (2001) “Occupational Gender Segregation and Effect on Wages in

Brazil” XXIV General Population Conference, Session 38 Labour Force

Osorio, R. (2006), “Desigualdades raciais e de genero no servicio publico civil” Cadernos

GRPE 2, 2006

Osorio, R. (2008) “A desigualdade racial de renda no Brasil: 1976-2006” Universidade de

Brasília: Instituto de Ciências Sociais, Departamento de Sociologia, PhD thesis

Paes de Barros, R., and Corseuil, C. H. (2004) “The Impact of Regulations on Brazilian

Labor Market Performance”, IADB Working Papers 3124

Pochmann M. (2007) "Nova geoeconomia do emprego no Brasil: um balanço de 15 anos

nos estados da federação", UNICAMP, mimeo

Queneau, H. (2009) “Trends in occupational segregation by race and ethnicity in the

USA: Evidence from detailed data” Applied Economics Letters 16: 1347-1350

Ramos, L., and Ferreira, V. (2005) “Padrões Espacial e Setorial da Evolução da

Informalidade no Brazil – 1991-2003” IPEA Texto para Discussao No. 1099

Reardon, S. F. and Firebaugh, G. (2002) “Response: Segregation and Social Distance – A

generalized approach to segregation measurement” Sociological Methodology 32: 85-101

Reilly, B. (1991) “Occupational Segregation and Selectivity Bias in Occupational Wage

Equations: An Empirical Analysis Using Irish Data” Applied Economics 23(1): 1-7

Sastre, M., and Trannoy, A. (2002) “Shapley inequality decomposition by factor

components: Some methodological issues” Journal of Economics 9, 51-89

39  

Semyonov, M., and Jones, F. (1999) “Dimensions of Gender Occupational

Differentiation in Segregation and Inequality: A Cross-National Analysis” Social Indicators

Research 46: 225-247

Shorrocks, A. (1999) “Decomposition procedures for distributional analysis: A unified

framework based on the Shapley value” University of Essex, mimeo

Silber, J. (1989a) “On the measurement of employment segregation” Economics Letters

30: 237-43

Silber, J. (1989b) “Factor components, population subgroups and the computation of the

Gini Index of inequality” Review of Economics and Statistics 71: 107-115

Soares, F. (2004) “Some Stylized Facts of the Informal Sector in Brazil in the Last Two

Decades” IPEA Texto para Discussão No. 1020

Soares, S. (2000) “O Perfil da Discriminação no Mercado de Trabalho – Homens Negros,

Mulheres Brancas e Mulheres Negras” IPEA Texto para Discussao No. 769

Soares, S., and Izaki, R. S. (2002) “A Participacão Feminina no Mercado de Trabalho”

IPEA Texto Para Discussão No. 023

Telles, E. (1992) “Who Gets Formal Sector Jobs? Determinants of Formal-Informal

Participation in Brazilian Metropolitan Areas” Works and Occupations 19(2): 108-27

Telles, E. (1994) “Industrialization and Racial Inequality in Employment: The Brazilian

Example” American Sociological Review 59(1): 46-63

Telles, E., and Lim, N. (1998) “Does it Matter Who Answer the Race Question? Racial

Classification and Income Inequality in Brazil” Demography 35(4): 465-474

Tomaskovic-Devey, D., Zimmer, C., Stainback, K., Robinson, C., Taylor, T., and

McTague, T. (2006) “Documenting desegregation: Segregation in American workplaces by race,

ethnicity, and sex, 1966-2003” American Sociological Review 71: 565-588

Tzannatos, Z. (1999) “Women and Labor Market Changes in the Global Economy:

Growth Helps, Inequalities Hurt and Public Policy Matters” World Development 27: 551-569

Ulyssea, G. (2006) “Informalidade no Mercado de Trabalho Brasileiro: uma Resenha da

Literatura” Revista de Economia Política 26, 4 (104): 596-618

Ulyssea, G. (2010) "Regulation of entry, labor market institutions and the informal

sector" Journal of Development Economics 91(1): 87-99

40  

Urani, A. (1996) “Ajuste macroeconomico e flexibilidade do mercado de trabalho no

Brasil – 1981/95” IPEA Texto para discussao No. 380

Wajman, S., and Rios Neto, E. L. G. (2000) “Quantas serão as Mulheres: Cenario para a

Actividade Feminina” in Maria Isabel Baltar da Rocha, Trabalho e gênero 59-84

Wajnman S., Oliveira, A. M., and Oliveira, E. L. (2006) “Older Brazilian in the Labour

Markets: Tendencies and Consequences, in

Watts, M. (1998) “Occupational Gender Segregation: Index Measurement and

Econometric Modeling” Demography 35(4): 489-496

World Bank (2002a), “Brazil Jobs Report. Volume I: Policy Briefing”, A Joint Report by

the World Bank and IPEA

World Bank (2002b), “Brazil Jobs Report. Volume II: Background Papers”, A Joint

Report by the World Bank and IPEA

41  

Table and Figures to be inserted in the text Figure 1: Shares of formal and non-formal sectors over time - as percentage of total labour force

Source: Author’s own computations using PNAD 1987 - 2006. Note: 1991, 1994 and 2001 missing years. Figure 2: Female and non-white shares across all labour markets over time

Source: Author’s own computations using PNAD 1987 - 2006. Note: 1991, 1994 and 2001 missing years.  

.2.3

.4.5

.6

1985 1990 1995 2000 2005year

All labour market Formal sectorInformal sector Self-employed sector

Female shares across sectors

.2.3

.4.5

.6

1985 1990 1995 2000 2005year

All labour market Formal sectorInformal sector Self-employed sector

Non-white shares across sectors

42  

Figure 3: The gender/racial composition across sectors

Source: Author’s own computations using PNAD 1987 and 2006.  

43  

Figure 4: Distribution of female and non-white workers in occupations Panel A – Female workers

Panel B – Non-white workers

Source: Author’s own computations using PNAD 1987 and 2006. Note: Panels A and B capture the number of women and men, respectively employed in each of the top five most numerous occupations, disaggregated across the formal, informal and self-employed sectors.

44  

Figure 5: Indices of segregation by gender and race over time Panel A – Duncan and Duncan index

Panel B – Karmel and Maclachlan index

Panel C – Gini segregation index

Source: Author’s own computations using PNAD 1987 - 2006. Note: 1991, 1994 and 2001 missing years.  

.4.5

.6.7

.8du

ncan

inde

x

1985 1990 1995 2000 2005year

All labour market Formal sectorInformal sector Self-employed sector

Duncan index by gender

0.1

.2.3

.4du

ncan

inde

x

1985 1990 1995 2000 2005year

All labour market Formal sectorInformal sector Self-employed sector

Duncan index by race

.2.3

.4.5

.6km

ind

ex

1985 1990 1995 2000 2005year

All labour market Formal sectorInformal sector Self-employed sector

Karmel & Maclachlan index by gender0

.1.2

.3.4

km in

dex

1985 1990 1995 2000 2005year

All labour market Formal sectorInformal sector Self-employed sector

Karmel & Maclachlan index by race

.5.6

.7.8

.9gi

ni in

dex

1985 1990 1995 2000 2005year

All labour market Formal sectorInformal sector Self-employed sector

Gini index by gender

0.1

.2.3

.4gi

ni in

dex

1985 1990 1995 2000 2005year

All labour market Formal sectorInformal sector Self-employed sector

Gini index by race

45  

Figure 6: Occupational segregation disaggregated by characteristics of the labour force Panel A – Demographic patterns

Panel B- Educational patterns

Panel C – Sectoral patterns (three main economic sectors)

Source: Author’s own computations using PNAD 1987 - 2006. Note: 1991, 1994 and 2001 missing years.  

.4.5

.6.7

Du

ncan

an

d D

unca

n in

dex

1985 1990 1995 2000 2005year

Young AdultElderly

Duncan index by gender

0.1

.2.3

Du

ncan

an

d D

unca

n in

dex

1985 1990 1995 2000 2005year

Young AdultElderly

Duncan index by race

.4.5

.6.7

Du

ncan

an

d D

unca

n in

dex

1985 1990 1995 2000 2005year

Illiterate Compulsory schoolMore than compulsory

Duncan index by gender0

.1.2

.3D

unc

an a

nd

Dun

can

inde

x

1985 1990 1995 2000 2005year

Illiterate Compulsory schoolMore than compulsory

Duncan index by race

0.1

.2.3

.4.5

.6D

unc

an a

nd

Dun

can

inde

x

1985 1990 1995 2000 2005year

Primary sector Secondary sectorTertiary sector

Duncan index by gender

0.1

.2.3

.4D

unc

an a

nd

Dun

can

inde

x

1985 1990 1995 2000 2005year

Primary sector Secondary sectorTertiary sector

Duncan index by race

46  

Panel D – Spatial patterns (five main regions)

Panel E – Spatial patterns (urban/rural areas)

Source: Author’s own computations using PNAD 1987 - 2006. Note: 1991, 1994 and 2001 missing years.  

.4.5

.6.7

Du

ncan

an

d D

unca

n in

dex

1985 1990 1995 2000 2005year

North North-EastSouth-East SouthCentral-West

Duncan index by gender

0.1

.2.3

Du

ncan

an

d D

unca

n in

dex

1985 1990 1995 2000 2005year

North North-EastSouth-East SouthCentral-West

Duncan index by race

.4.5

.6.7

Du

ncan

an

d D

unca

n in

dex

1985 1990 1995 2000 2005year

Urban areas Rural areas

Duncan index by gender

0.1

.2.3

Du

ncan

an

d D

unca

n in

dex

1985 1990 1995 2000 2005year

Urban areas Rural areas

Duncan index by race

47  

Figure 7: Contribution of different components to declining segregation (%) Panel A – Gender segregation

Source: Author’s own computations using PNAD 1987,1988, 1989, 1990, 1992 and 2002, 2003, 2004, 2005, 2006. Panel B – Racial segregation

Source: Author’s own computations using PNAD 1987,1988, 1989, 1990, 1992 and 2002, 2003, 2004, 2005, 2006.

48  

Table 1: Indices of segregation

All labour market Formal sector Informal sector Self-employed sector

1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006

Duncan index

gender 0.605 0.612 0.601 0.571 0.565 0.556 0.577 0.572 0.532 0.513 0.692 0.729 0.710 0.648 0.653 0.647 0.620 0.605 0.624 0.617

s.e. 0.002 0.002 0.002 0.002 0.002 0.004 0.004 0.003 0.003 0.003 0.004 0.004 0.004 0.003 0.004 0.004 0.004 0.004 0.004 0.004

race 0.199 0.200 0.199 0.195 0.191 0.193 0.192 0.195 0.175 0.177 0.197 0.153 0.163 0.183 0.191 0.134 0.153 0.160 0.153 0.149

s.e. 0.003 0.003 0.003 0.003 0.002 0.004 0.004 0.004 0.004 0.004 0.005 0.005 0.005 0.005 0.005 0.006 0.006 0.006 0.005 0.005

Karmel & Maclachlan index

gender 0.277 0.286 0.285 0.275 0.275 0.229 0.268 0.270 0.258 0.249 0.328 0.345 0.341 0.319 0.324 0.270 0.255 0.244 0.267 0.272

s.e. 0.002 0.002 0.002 0.001 0.001 0.002 0.005 0.002 0.002 0.002 0.004 0.004 0.004 0.003 0.002 0.005 0.007 0.004 0.003 0.003

race 0.099 0.099 0.099 0.098 0.095 0.093 0.092 0.095 0.086 0.088 0.097 0.074 0.080 0.089 0.091 0.066 0.076 0.079 0.076 0.073

s.e. 0.002 0.001 0.001 0.001 0.001 0.002 0.002 0.002 0.002 0.002 0.003 0.003 0.002 0.002 0.002 0.003 0.003 0.003 0.003 0.002

Gini segregation index

gender 0.776 0.785 0.775 0.743 0.735 0.713 0.738 0.725 0.701 0.678 0.819 0.845 0.838 0.810 0.810 0.823 0.794 0.764 0.756 0.759

s.e. 0.003 0.002 0.004 0.002 0.002 0.005 0.004 0.003 0.003 0.003 0.006 0.005 0.005 0.003 0.003 0.004 0.007 0.004 0.004 0.003

race 0.265 0.267 0.261 0.263 0.262 0.259 0.262 0.259 0.247 0.244 0.249 0.203 0.215 0.239 0.256 0.196 0.224 0.218 0.221 0.218

s.e. 0.003 0.003 0.003 0.003 0.003 0.005 0.005 0.005 0.004 0.004 0.006 0.007 0.006 0.006 0.006 0.007 0.006 0.006 0.006 0.006Source: Author’s own computations using PNAD 1987 – 1992 – 1997 – 2002 – 2006. Note: Standard errors boostrapped with 500 replications.  

49  

Table 2: Shapley decomposition of changes in Duncan index over time across sectors

Value of the index in the initial period

Value of the index in the final period

Changed observed between the two periods

Component of the change due to variations in the "internal structure"

Component of the change due to variations in the "margins" (1) + (2)

Component due to variations in the occupational structure (1)

Component due to variations in the shares of the sub-population (2)

All labour market

gender 0.5983 0.5624 -0.0359 -0.0343 -0.0016n.s. -0.0080 0.0064

s.e. 0.0011 0.0009 0.0014 0.0035 0.0032 0.0032 0.0006

100.00% 95.72% 4.28% 22.89% -18.60%

race 0.1981 0.1928 -0.0053 -0.0262 0.0209 0.0199 0.0010

s.e. 0.0014 0.0011 0.0017 0.0043 0.0041 0.0041 0.0001

100.00% 493.30% -393.30% -374.41% -18.89%

Formal sector

gender 0.5526 0.5167 -0.0359 -0.0367 0.0008 n.s. -0.0105 0.0113

s.e. 0.0018 0.0014 0.0023 0.0054 0.0053 0.0055 0.0012

100.00% 102.25% -2.25% 29.14% -31.38%

race 0.1891 0.1768 -0.0123 -0.0096 n.s. -0.0027 n.s. -0.0032 n.s. 0.0004

s.e. 0.0021 0.0017 0.0026 0.0065 0.0058 0.0058 0.0001

100.00% 77.71% 22.29% 25.80% -3.51%

Informal sector

gender 0.6949 0.6492 -0.0458 -0.0896 0.0438 0.0551 -0.0113

s.e. 0.0020 0.0016 0.0025 0.0103 0.0101 0.0113 0.0021

100.00% 195.66% -95.66% -120.29% 24.63%

race 0.1985 0.1835 -0.0150 -0.0604 0.0454 0.0423 0.0030

s.e. 0.0023 0.0020 0.0030 0.0083 0.0078 0.0077 0.0002

100.00% 401.82% -301.82% -281.75% -20.07%

Self-employed sector

gender 0.6300 0.6099 -0.0201 -0.0452 0.0251 0.0255 -0.0004 n.s.

s.e. 0.0020 0.0019 0.0027 0.0145 0.0143 0.0143 0.0005

100.00% 225.12% -125.12% -126.98% 1.85%

race 0.1294 0.1400 0.0106 -0.0266 0.0372 0.0359 0.0013

s.e. 0.0027 0.0022 0.0034 0.0089 0.0083 0.0083 0.0002

100.00% -249.96% 349.96% 338.11% 11.85%Source: Author’s own computations using PNAD 1987,1988, 1989, 1990, 1992 and 2002, 2003, 2004, 2005, 2006. Note: the initial period comprises 1987-1988-1989-1990-1992 and the final period comprises 2002-2003-2004-2005-2006. Standard errors boostrapped with 500 replications. n.s. indicates those components that are not statistically significant at 5% confidence.    

50  

Table 3: Shapley decomposition of changes in Duncan index over time disaggregated by characteristics

Value of the index in the initial period

Value of the index in the final period

Changed observed between the two periods

Component of the change due to variations in the "internal structure"

Component of the change due to variations in the "margins" (1) + (2)

Component due to variations in the occupational structure (1)

Component due to variations in the shares of the sub-population (2)

By age

young

gender 0.5791 0.5500 -0.0291 -0.0324 0.0033 n.s. -0.0010 n.s. 0.0043

s.e. 0.0017 0.0015 0.0023 0.0064 0.0063 0.0061 0.0007

100.00% 111.16% -11.16% 3.52% -14.68%

race 0.2019 0.2157 0.0138 -0.0205 0.0343 0.0337 0.0006

s.e. 0.0021 0.0019 0.0028 0.0076 0.0072 0.0072 0.0001

100.00% -148.09% 248.09% 243.86% 4.24%

adult

gender 0.6354 0.5926 -0.0427 -0.0411 -0.0016 n.s. -0.0067 n.s. 0.0051

s.e. 0.0017 0.0013 0.0021 0.0046 0.0043 0.0049 0.0021

100.00% 96.21% 3.79% 15.71% -11.92%

race 0.1995 0.1854 -0.0141 -0.0288 0.0148 0.0133 0.0014

s.e. 0.0022 0.0017 0.0027 0.0059 0.0053 0.0053 0.0001

100.00% 204.88% -104.88% -94.75% -10.14%

elderly

gender 0.6171 0.6002 -0.0169 -0.0610 0.0441 0.0430 0.0011

s.e. 0.0035 0.0024 0.0045 0.0107 0.0104 0.0105 0.0021

100.00% 361.70% -261.70% -255.04% -6.65%

race 0.1887 0.1839 -0.0048 -0.0494 0.0446 0.0437 0.0009

s.e. 0.0037 0.0029 0.0048 0.0108 0.0100 0.0100 0.0001

100.00% 1029.06% -929.06% -909.65% -19.41%

By education

illiterate

gender 0.6419 0.6139 -0.0280 -0.1177 0.0897 0.0815 0.0082

s.e. 0.0030 0.0041 0.0048 0.0220 0.0221 0.0224 0.0016

100.00% 420.45% -320.45% -291.08% -29.37%

race 0.0767 0.0939 0.0172 -0.0001 n.s. 0.0172 n.s. 0.0172 n.s. 0.0001 n.s.

s.e. 0.0034 0.0047 0.0059 0.0141 0.0137 0.0137 0.0001

100.00% -0.34% 100.34% 100.01% 0.32%

compulsory school

gender 0.6295 0.6275 -0.0020 n.s. -0.0314 0.0294 0.0249 0.0045

s.e. 0.0014 0.0012 0.0018 0.0067 0.0064 0.0065 0.0006

100.00% 1577.01% -1477.01% -1249.55% -227.45%

race 0.1400 0.1283 -0.0117 -0.0144 n.s. 0.0027 n.s. 0.0018 n.s. 0.0009

s.e. 0.0017 0.0015 0.0023 0.0088 0.0085 0.0085 0.0001

100.00% 123.03% -23.03% -15.31% -7.72%

more than compulsory school

gender 0.4438 0.3810 -0.0628 -0.0869 0.0241 0.0274 -0.0033

s.e. 0.0041 0.0028 0.0050 0.0123 0.0113 0.0113 0.0007

51  

100.00% 138.37% -38.37% -43.60% 5.24%

race 0.1747 0.2178 0.0432 -0.0381 0.0812 0.0897 -0.0084

s.e. 0.0051 0.0032 0.0058 0.0110 0.0106 0.0108 0.0011

100.00% -88.24% 188.24% 207.74% -19.50%

By main economic sectors

primary sector

gender 0.2419 0.0818 -0.1601 -0.0343 n.s. -0.1258 -0.1297 0.0039

s.e. 0.0050 0.0056 0.0074 0.0336 0.0337 0.0341 0.0015

100.00% 21.41% 78.59% 81.03% -2.44%

race 0.1429 0.1666 0.0238 -0.0027 n.s. 0.0265 n.s. 0.0262 n.s. 0.0003

s.e. 0.0036 0.0036 0.0052 0.0344 0.0340 0.0340 0.0001

100.00% -11.48% 111.48% 110.27% 1.21%

secondary sector

gender 0.5156 0.6542 0.1386 -0.0145 n.s. 0.1531 0.1526 0.0005

s.e. 0.0023 0.0018 0.0030 0.0086 0.0084 0.0085 0.0002

100.00% -10.47% 110.47% 110.08% 0.39%

race 0.1779 0.1911 0.0132 -0.0005 n.s. 0.0137 n.s. 0.0124 n.s. 0.0012

s.e. 0.0024 0.0024 0.0033 0.0113 0.0108 0.0108 0.0002

100.00% -3.42% 103.42% 94.27% 9.14%

tertiary sector

gender 0.6291 0.4792 -0.1499 -0.0954 -0.0545 -0.0513 -0.0032

s.e. 0.0016 0.0012 0.0021 0.0065 0.0065 0.0064 0.0003

100.00% 63.63% 36.37% 34.22% 2.15%

race 0.1977 0.1868 -0.0109 -0.0377 0.0268 0.0263 0.0006

s.e. 0.0020 0.0013 0.0024 0.0078 0.0077 0.0077 0.0001

100.00% 346.79% -246.79% -241.53% -5.26%

By rural and urban areas

rural areas

gender 0.6222 0.6171 -0.0051 -0.0648 0.0597 0.0842 -0.0244

s.e. 0.0033 0.0027 0.0042 0.0174 0.0172 0.0166 0.0019

100.00% 1275.74% -1175.74% -1656.87% 481.13%

race 0.1429 0.1532 0.0103 -0.0020 n.s. 0.0123 n.s. 0.0126 n.s. -0.0003 n.s.

s.e. 0.0033 0.0030 0.0047 0.0178 0.0172 0.0173 0.0003

100.00% -19.40% 119.40% 122.34% -2.93%

urban areas

gender 0.5807 0.5475 -0.0332 -0.0378 0.0046 n.s. 0.0003 n.s. 0.0043

s.e. 0.0014 0.0010 0.0017 0.0034 0.0032 0.0032 0.0007

100.00% 113.81% -13.81% -0.82% -12.99%

race 0.2021 0.1918 -0.0104 -0.0285 0.0181 0.0172 0.0010

s.e. 0.0016 0.0012 0.0020 0.0044 0.0040 0.0040 0.0001

100.00% 274.91% -174.91% -165.41% -9.50%

Source: Author’s own computations using PNAD 1987,1988, 1989, 1990, 1992 and 2002, 2003, 2004, 2005, 2006. Note: the initial period comprises 1987-1988-1989-1990-1992 and the final period comprises 2002-2003-2004-2005-2006. Standard errors boostrapped with 500 replications. n.s. indicates those components that are not statistically significant at 5% confidence.

52  

Appendix Table A1: Classification of occupational codes ISCO88 MAJOR, SUB-MAJOR, MINOR AND UNIT GROUPS

MAJOR GROUP 1: LEGISLATORS, SENIOR OFFICIALS AND MANAGERS

11 Legislators and senior officials

111 Legislators

112 Senior government officials

113 Traditional chiefs and heads of villages

114 Senior officials of special-interest organisations

12 Corporate managers

121 Directors and chief executives

122 Production and operations department managers

123 Other department managers

13 General managers

131 Managers of small enterprises

MAJOR GROUP 2: PROFESSIONALS

21 Physical, mathematical and engineering science professionals

211 Physicists, chemists and related professionals

212 Mathematicians, statisticians and related professionals

213 Computing professionals

214 Architects, engineers and related professionals

22 Life science and health professionals

221 Life science professionals

222 Health professionals (except nursing)

223 Nursing and midwifery professionals

23 Teaching professionals

231 College, university and higher education teaching professionals

232 Secondary education teaching professionals

233 Primary and pre-primary education teaching professionals

234 Special education teaching professionals

235 Other teaching professionals

24 Other professionals

241 Business professionals

242 Legal professionals (Lawyers and Judges)

243 Archivists, librarians and related information professionals

244 Social science and related professionals

245 Writers and creative or performing artists

246 Religious professionals

247 Public service administrative professionals

MAJOR GROUP 3: TECHNICIANS AND ASSOCIATE PROFESSIONALS

31 Physical and engineering science associate professionals

311 Physical and engineering science technicians

312 Computer associate professionals

313 Optical and electronic equipment operators

314 Ship and aircraft controllers and technicians

315 Safety and quality inspectors

32 Life science and health associate professionals

321 Life science technicians and related associate professional

322 Modern health associate professionals

323 Nursing and midwifery associate professionals

324 Traditional medicine praticioners and faith healers

53  

33 Teaching associate professionals

331 Primary education teaching associate professionals

332 Pre-primary education teaching associate professionals

333 Special education teaching associate professionals

334 Other teaching associate professionals

34 Other associate professionals

341 Finance and sales associate professionals

342 Business services agents and trade brokers

343 Administrative associate professionals

344 Customs, tax and related government associate professionals

345 Police inspectors and detectives

346 Social work associate professionals

347 Artistic, entertainment and sports associate professionals

348 Religious associate professionals

MAJOR GROUP 4: CLERKS

41 Office clerks

411 Secretaries and keyboard-operating clerks

412 Numerical clerks

413 Material-recording and transport clerks

414 Library, mail and related clerks

419 Other office clerks

42 Customer services clerks

421 Cashiers, tellers and related clerks

422 Client information clerks

MAJOR GROUP 5: SERVICE WORKERS AND SHOP AND MARKET SALES WORKERS

51 Personal and protective services workers

511 Travel attendants and related workers

512 Housekeeping and restaurant services workers

514 Other personal services workers

515 Astrologers, fortune tellers and related workers

516 Protective services workers

52 Models, salespersons and demonstrators

521 Fashion and other models

522 Shop salespersons and demonstrators

523 Stall and market salespersons and demonstrators

MAJOR GROUP 6: SKILLED AGRICULTURAL AND FISHERY WORKERS

61 Skilled agricultural and fishery workers

611 Market gardeners and crop growers

612 Agricultural workers – self-employed excluded

613 Crop and animal producers – self-employed

614 Forestry and related workers

615 Fishery workers, hunters and trappers

MAJOR GROUP 7: CRAFT AND RELATED TRADES WORKERS

71 Extraction and building trades workers

711 Miners, shotfirers, stone cutters and carvers

712 Building frame and related trades workers

713 Building finishers and related trades workers

714 Painters, building structure cleaners and related trades workers

72 Metal, machinery and related trades workers

721 Metal molders, welders, sheet-metal workers, structural-metal preparers, and related trades workers

722 Blacksmiths, tool-makers and related trades workers

723 Machinery mechanics and fitters

724 Electrical and electronic equipment mechanics and fitters

54  

73 Precision, handicraft, craft printing and related trades workers

731 Precision workers in metal and related materials

732 Potters, glass-makers and related trades workers

733 Handicraft workers in wood, textile, leather and related materials

734 Craft printing and related trades workers

74 Other craft and related trades workers

741 Food processing and related trades workers

742 Wood treaters, cabinet-makers and related trades workers

743 Textile, garment and related trades workers

744 Pelt, leather and shoemaking trades workers

MAJOR GROUP 8: PLANT AND MACHINE OPERATORS AND ASSEMBLERS

81 Stationary plant and related operators

811 Mining and mineral-processing-plant operators

812 Metal-processing plant operators

813 Glass, ceramics and related plant operators

814 Wood-processing- and papermaking-plant operators

815 Chemical-processing-plant operators (and machine operators)

816 Power-production and related plant operators (and machine operators)

817 Industrial robot operators

82 Machine operators and assemblers

821 Metal- and mineral-products machine operators

822 Chemical-products machine operators

823 Rubber- and plastic-products machine operators

824 Wood-products machine operators

825 Printing-, binding- and paper-products machine operators

826 Textile-, fur- and leather-products machine operators

827 Food and related products machine operators

828 Assemblers

829 Other machine operators not elsewhere classified

83 Drivers and mobile plant operators

831 Locomotive engine drivers and related workers

832 Motor vehicle drivers

833 Agricultural and other mobile plant operators

834 Ships' deck crews and related workers

MAJOR GROUP 9: ELEMENTARY OCCUPATIONS

91 Sales and services elementary occupations

911 Street vendors and related workers

916 Garbage collectors and related labourers

92 Agricultural, fishery and related labourers

921 Agricultural, fishery and related labourers

93 Labourers in mining, construction, manufacturing and transport

931 Mining and construction labourers

932 Manufacturing labourers

933 Transport labourers and freight handlers

MAJOR GROUP 0: ARMED FORCES

100 Armed forces

998 Mal definidas

999 Nao declarada

Source: Author’s own computations using PNADs. Note: The categories highlighted in yellow have been omitted because it was not possible to find the related occupations across all Brazilian datasets. 

55  

Table A2: Test of mean differences across sectors Duncan index - gender Duncan index - race

1987 1992 1997 2002 2006 1987 1992 1997 2002 2006

Formal vs Informal -23.688 -27.956 -26.329 -24.061 -29.268 Formal vs Informal -0.580n.s. 5.649 4.981 -1.316 n.s. -2.412

Formal vs Self-Empl. -15.109 -7.671 -5.994 -17.677 -20.439 Formal vs Self-Empl. 8.139 5.443 5.203 3.430 4.707

Informal vs Self-Empl. 7.559 18.254 18.217 4.631 6.827 Informal vs Self-Empl. 7.920 0.048 n.s. 0.518 n.s. 4.267 6.283

Karmel & Maclachlan index - gender Karmel & Maclachlan index - race

1987 1992 1997 2002 2006 1987 1992 1997 2002 2006

Formal vs Informal -19.498 -11.791 -15.621 -18.038 -23.908 Formal vs Informal -1.111 n.s. 5.448 4.899 -1.135 n.s. -0.947

Formal vs Self-Empl. -6.991 1.487 n.s. 6.110 -2.519 -5.961 Formal vs Self-Empl. 7.675 4.783 4.748 2.979 5.013

Informal vs Self-Empl. 8.374 10.894 17.366 12.412 13.408 Informal vs Self-Empl. 7.917 -0.401 n.s. 0.238 n.s. 3.654 5.277

Gini segregation index – gender Gini segregation index – race

1987 1992 1997 2002 2006 1987 1992 1997 2002 2006

Formal vs Informal -13.031 -16.286 -19.371 -26.223 -32.868 Formal vs Informal 1.262 n.s. 6.929 5.715 1.069 n.s. -1.775

Formal vs Self-Empl. -16.732 -7.177 -7.590 -11.459 -18.163 Formal vs Self-Empl. 7.135 4.533 5.249 3.531 3.828

Informal vs Self-Empl. -0.634 n.s. 5.870 11.788 11.368 11.652 Informal vs Self-Empl. 5.535 -2.280 -0.421 n.s. 2.153 4.848

Source: Author’s own computations using PNAD 1987 – 1992 – 1997 – 2002 – 2006. Note: z-tests reported in the table; n.s. not statistically significant; all other z-tests are statistically significant at 5% confidence.

56  

Table A3: Test of mean differences over time for each sector All labour market Formal sector Informal sector Self-employed sector

Duncan index - gender

1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006

1987 1987 1987 1987

1992 1.838 n.s. 1992 3.846 1992 6.357 1992 -4.257

1997 -1.225 n.s. -3.142 1997 3.075 -0.926 n.s. 1997 3.078 -3.373 1997 -6.881 -2.575

2002 -10.780 -12.585 -10.129 2002 -4.646 -9.172 -8.499 2002 -8.247 -15.129 -11.727 2002 -3.880 0.572 n.s. 3.279

2006 -12.861 -14.666 -12.344 -2.021 2006 -8.342 -13.221 -12.671 -4.006 2006 -7.202 -14.016 -10.629 1.055 n.s. 2006 -4.998 -0.547 n.s. 2.145 -1.176 n.s.

Duncan index - race

1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006

1987 1987 1987 1987

1992 0.205 n.s. 1992 -0.188 n.s. 1992 -5.820 1992 2.391

1997 -109.677 -0.262 n.s. 1997 0.330 n.s. 0.520 n.s. 1997 -4.616 1.384 n.s. 1997 3.210 0.819 n.s.

2002 -111.811 -1.164 n.s. -0.928 n.s. 2002 -3.211 -2.966 -3.668 2002 -1.962 4.100 2.793 2002 2.544 0.007 n.s. -0.867 n.s.

2006 -121.453 -2.242 -2.030 -1.053 n.s. 2006 -2.975 -2.722 -3.448 0.394 n.s. 2006 -0.904 n.s. 5.335 4.032 1.165 n.s. 2006 1.984 -0.597 n.s. -1.487 n.s. -0.650 n.s.

Karmel & Maclachlan index - gender

1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006

1987 1987 1987 1987

1992 3.557 1992 7.026 1992 2.789 1992 -1.663 n.s.

1997 3.058 -0.548 n.s. 1997 13.178 0.337 n.s. 1997 2.079 -0.779 n.s. 1997 -4.005 -1.416 n.s.

2002 -1.165 n.s. -5.291 -4.769 2002 9.203 -1.880 -4.242 2002 -1.759 n.s. -5.246 -4.446 2002 -0.414 n.s. 1.570 n.s. 4.786

2006 -1.228 n.s. -5.265 -4.750 -0.125 n.s. 2006 6.377 -3.489 -7.280 -3.028 2006 -0.891 n.s. -4.480 -3.631 1.314 n.s. 2006 0.246 n.s. 2.099 5.638 0.921 n.s.

Karmel & Maclachlan index – race

1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006

1987 1987 1987 1987

1992 0.209 n.s. 1992 -0.312 n.s. 1992 -6.263 1992 2.419

1997 0.045 n.s. -0.170 n.s. 1997 0.601 n.s. 0.894 n.s. 1997 -4.879 1.590 n.s. 1997 3.213 0.822 n.s.

2002 -0.713 n.s. -0.944 n.s. -0.783 n.s. 2002 -2.846 -2.416 -3.491 2002 -2.246 4.212 2.717 2002 2.665 0.154 n.s. -0.704 n.s.

2006 -1.997 -2.264 -2.119 -1.326 n.s. 2006 -2.023 -1.605 n.s. -2.690 0.986 n.s. 2006 -1.832 n.s. 4.842 3.322 0.506 n.s. 2006 2.000 -0.607 n.s. -1.490 n.s. -0.804 n.s.

   

57  

Gini segregation index - gender

1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006

1987 1987 1987 1987

1992 2.287 1992 3.746 1992 3.226 1992 -3.737

1997 -0.293 n.s. -2.288 1997 1.997 -2.471 1997 2.526 -0.869 1997 -10.750 -3.813

2002 -8.675 -14.855 -6.985 2002 -1.931 n.s. -7.499 -5.597 2002 -1.253 n.s. -5.706 -4.942 2002 -12.527 -4.881 -1.444 n.s.

2006 -10.757 -17.819 -8.717 -2.945 2006 -5.731 -12.373 -11.090 -5.579 2006 -1.306 n.s. -5.803 -5.041 -0.076 n.s. 2006 -12.671 -4.663 -1.031 n.s. 0.498 n.s.

Gini segregation index - race

1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006 1987 1992 1997 2002 2006

1987 1987 1987 1987

1992 0.497 n.s. 1992 0.347 n.s. 1992 -4.974 1992 2.923

1997 -0.721 n.s. -1.229 n.s. 1997 -0.023 n.s. -0.389 n.s. 1997 -3.859 1.283 n.s. 1997 2.362 -0.652 n.s.

2002 -0.491 n.s. -1.031 n.s. 0.277 n.s. 2002 -1.693 -2.077 -1.777 2002 -1.058 n.s. 4.035 2.862 2002 2.647 -0.365 n.s. 0.297 n.s.

2006 -0.731 n.s. -1.286 n.s. 0.053 n.s. -0.247 n.s. 2006 -2.232 -2.635 -2.364 -0.513 n.s. 2006 0.933 n.s. 6.151 5.035 2.075 2006 2.433 -0.700 n.s. -0.020 n.s. -0.331 n.s.

Source: Author’s own computations using PNAD 1987 – 1992 – 1997 – 2002 – 2006. Note: z-tests reported in the table; a not statistically significant; all other z-tests are statistically significant at 5% confidence.